Professional Documents
Culture Documents
II
LLU
DISTRIBUTION AND AVAILABILITY
ON BACK COVER
80 1 b001
( i,.ARD-C-28
I COMPOSITE MATERIALS
--- I
DISTRIBUTION STAT X1
Apmrood for pul!. c roloao;a
t11izirbivion U,' !V1L!Et d
Papers presented at the 50th Meeting of the AGARD Structures and Materials Panel,
held in Athens, Greece on 14-17 April 1980.
4T
THE MWSSION OF AGARD
:1 The mission of AGARD is to bring together the leading personalities of the NATO nations in the fields of science
and technology relating to aerospace for the following purposes:
- Continuously stimulating advances in the aerospace sciences relevant to strengthening the common defence
postule;
- Improving the co-operation among member nations in aerospace research and development;
- Providing scientific and technical advice and assistance to the North Atlantic Military Committee in the field
of aerospace research and development;
- Rendering scientific and technical assistance, as requested, to other NATO bodies and to member nations in
connectton with research and development problems in the aerospace field,
- Providing assistance to membei nations for the purpose of increasing their scientific and technical potential;
I Recommending effective ways for the member nations to use their research and development capabilities for
the common benefit of the NATO community.
The highest authority within AGARD is the National Delegates Board consisting of officially appointed senior
I representativc- from each member nation. The mission of AGARD is carried out through the Panels which are
composed of experts appointed by the National Delegates, thc Consultant and Exchange Programme and the Aerospace
Applications Studies Programme. The results of AGARD work are reported to the member nations and the NATO
S
I Authorities through the AGARD series of publications of which this is one.
Participation in AGARD activities is by invitation only and is normally limited to citizens of the NATO nations.
4=
.; - __ __ _ _tSBN__Printedby Technical Editing and __ _ __ __92-835-0273-6--
Reproduction Ltd - == - -=- -- -- =-----
Ilarford House, 7-9 C Aharlotte
St London,
18D WIP
RESUME DU PRESIDENT
par
Georges Juhb6
Sous-Directeur Technique
Adrospatiale
37 Boulevard de Montmorency
t 75781 Paris, Cedex 16
L'objectif anibitieux de cc Congr~s dtait de faire le point du niveau des connaissances sur le comportement des
matt~riaux composites soumis aux divers types d'agressions que l'on recouvre aujourd'hui du vocable d'...environnement".
On doit admettre que cette revue 6tait particul i~rement opportune au moment oai se situe la chae~ni~re entre la
pt~iode de dt~veloppement techinologique, dt~jA accomplie, et I'application en service dont de!, r~sultats significatifs sont
naintenant disponibles.
L ~On peut constater que cc Congr~s a mis successiveinent en evidence deux tendances que l'on pourrait, Apreinifre
Aupr~s de ces facteurs d'inquiftude. 1'expdrience en service se traduit d'une faqon assez contradictoire par la
satisfaction, l'optimisnie, l'absence dt; difficultds majeures et en Fia de compte une vigoureuse poussde en avant, dans
laquelle les utilisateurs e-nbo~tent le pas aux promoteurs.
11est difficile de ne pas etre tcnt6, dins cc dilemme, de rejeter les avertissements des scientifiques et de consacrer
le trioinphe de la pratique sur ]a speculation theoriquc.
La question nWest pas aussi simple, car les ph~nom~nes mis en dvidence zn laboratoire sant ind~niables. 11leur
manque simplement les donndes statistiques qui permettraient de d~finir leur degrd de probabilit6.
- ~ D'un autre cbt6, il est vraisemblable que les c6l6mentts mis en service avaient b~ndficid d'unc fabrication relativement
bien surveillde, effectudte dans des ateliers travaillant dans des conditions de fabrication prototype. Qu'en sera-t-il
lorsque Iecomposite sera produit en s~rie avec des ddfauts i la limite des conditions d'acceptation, lesquelles auront
~ bien dfi, entre temps, dtie quelque peu Iib~ralis~es?
~ La solution finale se trouve donc a la convergence des deux cheminements, lorsque les deux tcrndances se
rejoindront et se compliteront dans une appreciation raisonnde du risque.
Un nouveau rendez-vous est ddjA pris; entre-temps, tous les 6l6ments sont rdunis pour que le matdriaux composite
poursuive, sous survemiiance, sa croissanc,! et son d~veloppement.
Accession 'or
NTIS ('0 &
DTIC VB
Distribut4-0n/
AvailabiitY Codos
SIjA\ail and/or
_ _ __ _ _ Di~I Spec,1
iwl
CONTEN "S P
Pg
RESUME DU PRESIDENT
par Georges Jubd iii
Reference
~ --- iv
Referen xe
by A.J.Barnard \ 7
LEV
Eig
m
Oop 'p -
SUMMARY
The available evidence on the effects of long term natural weathering of carbon fibre composites is
examined and its implications with regard to the conditioning of test specimens discussed.
The effects of laboratory accelerated tests on the properties of some composite materials are
considered in conjunction with the relevance of da'a thus acquired to real life situations and the need to
generate the data with reasonable speed for design considerations.
The changes in conditioning procedure which have taken place are outlined along with the factors which
have influerced these changes.
1. INTRODUCTION
Since the adverse effect of absorbed moisture on the material properties of carbon fibre composites
was first noted, those concerned with the design of aircraft structures using these new materials have been
considering the problem of including a realistic assessment of this loss of property in the determination
of design allowables.
In military aircraft applications, which are the main consideration of this paper, the most severe
design condition is usually taken to be the "hot wet" case. A simple view of this condition is the simul-
taneous occurrence of the maximum moisture content anda high (often > 1000C) design temperature. Two
problems present themselves here. There is the need to assess the 2pvel of moisture content likely to be
txperienced in service life. (Discussion of temperature levels is outside the scope of this article).
There is also the need in the planning of experimental procedure to decide on the best method of achieving
this representative level of moisture, bearing in mind that the real life timescale is in most cases much
too slow (Q months - 3 years) to be capable of integration into the design process. Clearly laboratory
accelerated conditioning will continue to be used in the foreseeable future. Accepting this, in order to
prevent incorrect assessment of the experimental results thus obtained, it is necessary to assess how the
state of the material to be expected after such accelerated conditioning relates to that likely to be
encountered in service.
In this paper consideration is first given to the evidence now accumulating on a worldvide basis on
the moisture levels likely to be found in CFC components, and the effect which this natural weathering,
which includes among other things exposure to ultra-violet rays, thermal gradients bLtween surfaces exposed
and sheltered from sunlight, precipitation of various types, wind and sand, is expected to have on the
condition and strength and stiffness properties of the material.
In Section 3, three methods of accelerated conditioning which have communly been used are compared.
Attention is given to the total level of moisture uptake, the diatribution of moisture through the thickness
S [of the material, the relationship between moisture content and loss of mate-rial property and the mechanisms
of degradation occurring. An attempt is made to assess how these effects produced under these artificial
"hot house" conditions are representative of real life situations. In Section 4 consideration is given to
the need to know with some precision the actual moisture levels obtaining in test spec.mens and the evolution
Sof experimental practice _n this area within BritisL Aerospace is outlined.
Before it is possible to decide on the merits and demerits of the various possible forms of accelerated
laboratory conditioning, it is necessary to consider what level of moisture uptake should be the target.
This can only be fixed with confidence when the required amount of information has been gained from real
time weathering experiments. In addition to revealing the moisture content these tests will also provide
information on the property and state of the material to be expected in service. The timescale of these
experiments must of necessity be several years although useful intermediate results can be obtained by
periodic withdrawal of exposed material.
Although it can not yet be honestly stated that the available information is adequate, test results
are now beginning to come in and certain trends are appearing. Material haa been exposed on various sites
round the world, with measurements of moisture uptake, loss due to ultraviolet, and residual strength. In
Hangar 32 at British Aerospace, Warton, which houses Jaguar BO0, regular measurements of temperature and
humidity have been taken since July 1978.
Predictive calculations based on these measurements are illustrated in Fig. 1. These calculations are
compared with moisture uptakes deduced by dessicating coupon specimens cut from discs attached to the
aircraft, and these practical measurements are also given in Fig. 1. The actual flying time during the
period under consideration is small and has been ignored in the predictive calculations, but it should be
J . " t'
1-2
borne in mind that the effect of flight is genarally to dry the material, at least near the surface, and
this may account for some of the difference between the theoretical and practical results.
Roylance and Roylance(l) give r~sults of exposure at open field sites in Panama of specimens from a
3Y 1009-26 Scotchply glass/epoxy laminate, Sxposure was in an unloaded condition facing south and 45 deg.
to the vertical. As samples are returned for analysis increasing degradation of the outer surfaces,
especially those facing the sun, is noticed. Resin is lost from the surface, leaving behind exposed fibres.
Fig. P shcow the results of glass transition temperature measurements plotted against months of exposure.
The relatively undegraded state of the middle laminate layers should be noted.
Trabocco and Stander(P) give results of open field exposure in Pennsylvania and Panama. Degradation
was measured by tension, compression and short beam shear tests. Unfortunately none of these property
tests were at elevated temperature and no measurements of moisture uptake are available. However it must
be remembered that this and other real time exposures were programmed several years ago when the need for
such measurements had not been fully recognised. The exposed material included coated and uncoated panels,
and a number of different fibre/resin combinations which produced a considerable variety of behaviour. The
greatest effect was observed on material exposed in Panama. Modulite 5206 (Modmor II fibres) actually
appeared to benefit from 18 months tropical exposure, it being suggested that post-curing was stimulated
by the conditions. AS/3501 appeared to be little affected in mechanical propertiet.. In contrast T4OO/^344
panels shoyed pronounced warping, loose fibres and loss of resin as was the case with the glass-epoxy
material 1 . Additionally this material suffered considerable loss of tensile strength and ultimate st,_..
properties. This system, however, is not a contender for British aircraft applications. Fig. 3 and 4
show the loss of ultimate strain and tensile strength by T40C/2544. However, more typical are Fig. 5 and
6 showing negligible loss of compression strength by AS/3501 and 3002T after i1 months exposure at Warminster
(Pennsylvania) and Panama.
Pride(3) gives the results of experiments in which specimens have been mounted on racks which are
placed on rooftops of airline buildings at a number of locations round the world. Fig. 7 shows the weight
change with exposure time for a 4 -ply T300-5209 laminate exposed unstressed at Langley USAF. Vs.. The
specimens initially gained weight then began to lose weight under the influence of solar ultra-violet
exposure. After 140 days the exposure was terminated and the moisture content determined by dessication.
ThA ultra-violet loss is taken to be the difference between final dry weight and the initial pre-exposure
measurement. The ultra-violet loss is assumed to be linear with time and the curve of moisture uptake is
constructed on this basis. Fig. 8 summarises the results of moisture content measuremente on various
material combinations, these values being corrected for ultra-violet loss as previously described. Also
taken from this reference is Fig. 9 which compares measured moisture contents with those predicted with
and without allowance for solar effects. Residual mechanical property tests were conducted at ambient
only. These revealed some scatter but no significant change in strength.
There have been a number of papers in recent yaars which have given the results of calculations 9
moisture uptake based on appropriate meterological information. Typical of these is Unman and Tenney(4).
They used weather bureau data for Langley AFB and Norfolk, Vs. to calculate the amount of moisture a
T3OO/5?07 composite panel would contain if exposed outdoors. Solar radiation data along with cloud and
wind information were used. Results are quoted with and without the solar effect. Good agreement was
found between the calculated moisture content and nublished data for T,400/5208 composite. Fig. 10 and 11
are reproduced fror this reference. Fig. 10 illustrates the effect of solar radiation on the moisture
content of a 12-rly T300/520 horizontal ponel exposed at Norfolk Vs. Fig. 11 Fives the calculated moisture
contents for12-sy Ti200/52CP laminates using monthly average weather data for five different sites.
Fig. 12 shows the results of calculations performed at B.Ae Warton, without solar effects, for
A.A*P/ql4C material with an effective skin thickness cf 2rr.,
based on meterological information for Gemert
(Netherlands). Fig. 13 shows the through the thickness moisture distribution at the times of maximum and
minimum total concentration in the quasi-steady state.
Considering the available information there would appear to be little Justification for conditioning
test pieces to a greater level of moisture uptake than 11'when due allowance is made for solar and flight
d:yinr effects, indeed for temperate climates 0.8& might be nearer the mark. This level is however fairly
evenly distributed through the thickness. The damaging effect of ultra-violet in the surface layers, an
apparent feature of natural weathering, will not of course be reproduced in laboratory experiments, but
& as it is confined to the outermost layers the loss of structural integrity is expected to be small.
These exposures were used in a joint programme conducted by B.Ae rarton and the Royal Aircraft Establishment,
r Farnborough. Exposure (b) is now used at Warton as a "slow" chamber, while exposure at 950 C and 95-100% PR.H
is used as a "fast" chamber. The timescales of the design process largely dictate the use of the "fast"
exposire when ideally a somewh.at less severe conditioning would be preferred. The inclusion of exposure (c)
may cause some surprise, but despite its mild appearance the results were considered unexpectedly severe
at the time, and equilibrium roisture levels are apparently somewhat higher than those to be expected in
service.
In the joint RAE/B.Ae programme moisture ingress levels and ILSS were measured for the thrpe exposures
L_3
I,f on two fibre-resin combinations, HTS/BSL914C and AS/3501/5A. Results from this programme are used in this
31-3
comparison, but it emphasised that the inferences drawn from these results are those of the author, and do
not necessarily represent the views of the RAE or other parties. In carrying out this comparison, attention
has been concg-ntrated on three areas:
'i) Equilibrium concentration
Fig. 14 shows the equilibrium concentrations measured in or deduced from these experiments plotted
against relative humidity. It will be observed that the concentrations recorded in the boiling water
immersions are much greater than would be expected by extrapolating the relationshxp between concentration
and humidity obtaining at lower levels. This was also noted by McKague et.al. and is in harmony with
other results( 6 ) which show total immersion giving higher concentrations than exposure in a chamber to
maximum humidity.
Fig. 15 and 16 illustrate the relationship between moisture content and ILSS strength at 130C. This
featire is perhaps better shown in rable 1, which gives the results of a linear regression analysis performed
o..the test points. A linear best fit is obtained for the equation,
S = AM+ B
F = moisture content %
and the computed values of A and B are given in Table 1. A fair amount of scatter is evident but there
appears to be basically a similar relationship between IISS and moisture content at all exposure conditions.
This can be definitely seen in the case of HTS/914C laminate 2545.
Fig. 17 and 18 show the relationship between moisture content and ILSS strength at R.T. In this case
there is a somewhat different relationship than at 130 0 C. For exposure (c) no significant change is
apparent. For exposure (b), up to about the maximum level achieved with exposure (c), there is also no
si eni'icant change; at higher moisture levels there is a small, erratic loss of property. For exposure (a),
rciling water immersion, there is a drop of property at all levels, erratic for small uptakes, but more
pronounced at higher levels.
Unfortunately aot many tests where specimens, having beer exposed, were dried out and tested to
investigate recovery properties, have been performed after exposures (b) and (c). However, such as have
been carried out have indicated ccmplete or nearly complete recovery of ILSS at 130 0 C. Tables 2 and 3
show the results for exposure (a) and these generally show poor recovery of property.
This comparison suggests that exposure (a) leads to unrealistically severe specimen degradation, and
despite the interesting results obtained with it in the past it would appear, bearing in mind other
evidence(O) also, that when rapid results are required it is better to expose at a condition similar to
that of the Warton "fast" chamber (950C and 95% R.H.) rather than go to total immersion. Exposure (b)
would appear to give somewhat excessive degradation at higher values of moisture content, but only when
the expected maximum content in service has been exceeded. The timescale of exposure (c) is too slow to
be much used in the design process. Fig. 19 illustrates the theoretical through the thickness distributions
for each of the three exposures corresponding to the point where the total moisture content is 1.0%.
Comparing these to the expected real life situation (Fig. 13) reinforces the remarks just made about the
relative severity of the exposure conditions.
fI ith particular reference to the poor recovery properties, it would appear that the boiling water
exposure, in contrast to less severe tests, leads to a considerable degree of irreversible degradation.
The distinction between irreversible and reversible degradation is of course not clear cut. However
effects due to the plasticisation of the resin r.sulting from moist.ure content in the form of solution
appear to be practically completely reversible at least as far as static strength is concerned. This may
not be true to the same extent for stress-strain and fatigue. Micrucracking o. the resin due to moisture
and thermal gradients and ultra-violet exposure would in contrast appear to be largely irreversible. The
degradation of the fibre-resin interface is more complex. The bond between fibre and matrix appears to be
part chemical and part friction. A friction bond would be expected to weaken in the presence of moisture
and to reform on drying out provided the local fibre resin alignment had not been greatly distuibed.
However once chemical bonds are ruptured then at best it would appear that a much longer timescale would
be necessary for these to be reformed so that this type oa debonding can probably be classed as irreversible.
This may also be the case where water molecules appear to become permanently bound in the resin(?,O).
If it is accepted that apart from the ultra-violet effects on the surface layers, degradation to be
expected in service is if anything less severe than that observed under exposure (c), then this would
appear to imply that property loss will be largely reversible. Thus at the very least care will be
required in reading across from results where it is considered thai significant irreversible effects may
have occurred.
Until fairly recently testing for material property within British Aerospace(9, 10) was done for "dry"
and "wet" conditions without actual measurements of moisture content. It was at RAE(11) that the first
British experiments took place in which the moisture uptake was systematically monitored. To obtain the
1-4
"wet" condition specimena were given a standard conditioning of 2000 hours at 50 (C and 95-10% R.H. The
"dry" specimens were not dried to any predetermined datum. Thry were not given any conditioning and clearly
had absorbed moisture during the inevitable dwells between manufacture and test.
A combination of circumstances have led to a change of attitude. The RAE work and the earliest
results from real time exposure tests showed that the standard "wet" condition would give pessimistic
results from the design point of view, at least for the thinner specimens. The earliest &ttempts at
predictive calculations(12) drew attention to the need to give different exposures according to thickness,
to the rate of absorption in ambient conditions (typically more than a quarter at 200C of that at 5OOC) and
to the possible significant difference in through the thickness distribution between laboratory conditioned
specimens and real life situations.
These considerations have led to a change in procedure at B.Ae Warton, where the "dry" condition is now
obtained by drying test pieces to constant weight in an oven and the "wet" conlition is 0.9%-1.0% measured
moisture uptake. Simple predictive calculations are used to give guidance on exposure conditions and
timings, with attention to the evenness of uptake where necessary, as for instance in the case of fatigue
specimens. Fairiy sophisticated models are now used, particularly in the U.S.A., which include attempted
simulation of thermal and moisture cycling.
5. C0ONCLUSIONS
(a) Moisture uptakes to be expected in service life appear to be considerably lower than the
equilibrium concentration produced in CFC material by a high humidity envivonment.
(b) While it would obviously be desirable to reproduce real life rates of moisture absorption,
these are much too slow to be integrated into the design process.
(c) The excessive moisture contents produced by earlier standards of conditioning are now recognised,
but the danger of inducing excessive degradation by over-acceleration remains even when the actual moisture
level corresponds to real life expectations.
(d) Total immersion in water, particularly boiling water, does appear to give over severe effects,
with an unduly high proportion of irreversible damage and unrealistic through the thickness distributions.
(e) Increased knowledge and experience, allied with improvements in predictive techniques and
leading to a greater degree of control over the conditioning process, is reducing the danger of incorrect
design decisions arising from misinterpretation of experimental results.
6. REFERENCES
6. R. Long, Jr. "Moisture Diffusion Parameter Characteristics for Epoxy Composites and Nest
Resins". 1979. NASA Technical Paper 1474.
SComnosites." 9. J. P. Prestwich.
10,78. B.Ae Manchester Division.
HSA-MSF1,-R-G!24-0295. "Unidirectional Properties of Carbon Fibre
I. {10. J. Sastham. B.Ae Warton Division. "Final Report on M.O.D. Contract K/LR32B/216. In Depth
Evaluation of C.F.R.P."' 1979. MDR 01F2.
1'.C. W. Judd. RAE (Farnborough). "The Effect of Water on Carbon Fibre Composites. Pa3rt IV".
1173. PAE Tech. Memo. MAT 225.
12. J. Crank. "The Mathematics of Diffusion." ?nd Edition. 1975. Clarendon Press. Oxford
I-5
7. ACKNOWLEDGE14ENTS
Thanks are due to the Directors of British Aerospace for permission to publish this paper, and to the
Royal Aircraft Establishment, Farnborough, for permission to use experimental results from the joint B.Ae/
RAE programme. In his work on moisture absorption of CF, and related phenomena the author is grateful for
the encouragement of hr. T. Sharples, Technicel lManager, Advanced Structural Applications Department,
British Aerospace, Warton Division.
R
iI
V
.:.2
A
1-6
APPENDIX
TABLE 1
The coefficient A and B given in the table below are those given by the linear regression analysis
described in Section 3. Graphical presentation of some of the results is given in Fig. 15 and 16.
The values of A and B have been deduced for which the equation
S AM+ B
M = moisture content %
is a best fit to the test iesults for IISS at 130C, based on minimum r.m.s. deviation.
Material Exposure A B
TABLE 2
Table 2 following gives the results of ILSS tests conducted at 1300 C in the joint BAe/RAE programme
on AS/3501/5A specimens which had been immeitsed in boiling water then dried to constant weight in an oven
set at 500C.
Control 80.7 - - -
"Measured absorption during experiments - no allowance made for unmonitored absorption prior to
It -" immersion.
S1-7
TABLE -3
Table 3 following gives the results of ILbS tests conducted at 130 0 C in the Joint BAe/RAE progrsmrm-
on HTS/914C specimens from laminate 2545 which had been immersed in boiling water then dried to constant
weight in an area set at 50 0 C.
Control 60.0 - -
3 55.3 0.30 55.3 0.0
fil
v ___
1.2.
I'0.
0 0
. 0
00
PRAC
P R TIC AL MSA5URE ME TS
.K 0 0 oSs- PLY
/X UNIDIRCTI 0 NAL
i'
100 o00 0o0 400 o00 0o0
JULY AUra SEPPI OCT NOV DEC JAN FES MAIR APR' MAY JUNE JULY AUG SEPi OCT' NOV' DEC. JANY
"18 78 78 78 78 73 79 79 79 19 79 79 79 79 79 79 79 79 90
Fig.l Moisture uptake prediction for Jaguar B08 based on temperature and humidity
readings in Hanger 32 BAe Warton
ISO
o 170]
ISO
' D M 0" Oa
D
ISO-
0 0O
OAPEATSIT
LAE XOUETM,1O S(AAA
Fig.120 M's
S12. 4
{ [ Fig.2 Measured glass transition temperatures after real time exposure (Ref.]I, Fig.6)
1-9
MO0N TH S O F E X P0SU R E
C2 400C 0
_3 300,
100
0~ '42
MOTH OFEPSR
2 400.
-I
100'
0 4 8 10 It 14 Ir jg
MONTHS 0 EXPOSUK.
SO0-., -" ,-
0~
F 4500
-
H
H O----Q COATEO
C20 WAR EXPOSURES
Stoo.
I 04 6 9 1 1'2 14 16 is to
Lii :
i
"':
Fig.6
m
MONTHS
(Ref.2, Fig.3)
oF EXPOS URE
MESUE M 0/ T
4~~U EFU1EFEC.
EX OS R T IES
MLE 5 0tAUY
(Rf3 Fig 10
KI E -F1
K4 tf1
T3 0.-2
E XiPOS RU IM , Y AR
Fi. osuepcu fe oldieepsrs(e.,Fg 3
I 4 -- - -~~
t7
1-12
I. S
S27 HO SOLAR
) SOLAR
.4
j ~OAT WORLDWIDE
A EXPOSURE
0 I
SEXPOSURE T4
I-. 1
I
NO SOLAR
i Io
.-. 0
SM. TA'HPA ,
GUA 0 0
7 10-2
WA SHINGTON D.C. W
o 30z
ce HoI
II
7 - I-o
o / LAS VEGAS
-J
1202
EXPOSURE. YEARSNEE
NO S0OLAR
n (..t
-. I T
11'AMS Nxowu
1-14
1.3
.. uN 0 S 0 L A Ru
1.0
4 .. 1.0
DISTANCECFROM EXPOSED SURFACE (mm.)
Fig.13 Through the thickness moisture uptake prediction for RAF Bruggen based on information
for Gemert (Netherlands). Predictions for times of maximum and minimum total moisture
contwat. Equivalent skin thickness 2mm
In
70"0 s o 0 I00
RE LA 71V E
R HUMIDITY - "to
40-.
-'A
100,
S C 95%~ R.I4.
EXPOSURE
[to
I 0-
In
_-
1100
'to
'I
70 _ _ _ _ _______
1-17A
Ii!
Ffa
j*~X
0
SOILING
SC 9&*/.
WATER IMMERSION
V LKEPOSURE
144
- --- --- - 0* M p
2-I
R. C. Tennyson
University of Toronto, Institute for Aerospace Studies
4925 Dufferin Street
Downsview, MOntario, Canada
5T6
SUMMARY
This report presents experimental results obtained on the chort and long term behaviour of polymer
matrix composites subjected to various environmental conditisna. Changes in mechanical stiffness,
strength and cc.fficiept of thermal expansion were measured under ambient pressure and thermal vacuum
conditions (10-0 - 0-o torr). It should be noted that in all beats involving the vacuum environment,
rwasurements were made in-situ, thus necessitating the use ot mechanical loading fixtures acting through
flexible bellows to pravide stiffness and strength data. Results are also given on the effects of
varying some fabrication parameters (length of post-cure time and rate of cool-down in autoclave) on the
changes observed in strength and stiffness. The materials investigated include: graphite/..poxy, boron/
epoxy, PRD-49/epoxy and E glass/epoxy.
Ih NOMENCLATURE
(Q11 - Ql~
1 2 -263 n +(, 22+26))
I Q1 El/(l - lv12va)
0 Q616
S,
12 S'
v22
2 /('- v v~l)
12
t T temperature
X,X' tensile and compressive lanina strengths measured in 1 direction
Y, Y' tensile and ccpressive lamina. strengths measured in 2 direction
: L laminate
0i midplane value
2-2
1, 2 lamina material axes parallel asod orthogonal to fiber (or filament) reinforcement
Greek Symbols
a coefficient of thermal expansion
7 sh;ar strain
a normal strain
K curvature
V Poisson's ra+io
f normal stress
shear stress
1. 1TRODUCTION
Aerospace applications of composite materials are increasing every year. One of the major conceims
in the use of composites, particularly for primary structure and hardware components, is the effe-t of
varicvu -nvironmental conditions on the material properties. This is especially true for polymer matrix
syvtems which can be quite sensitive to temperature, moisture aia radiation for example. Although in
( ce n extensive environmental test data can be found in the lite'ture, a scarcity of information
exiats on space environment effects.
During the life cycle of a structure that must operate fall or part-time in space, it will experi-
ence an exceedirgly !Liarsh -nvironment composed of temperature ext,'emes, ultra hard vacuum, a broad
spectrum of intense radiation, micro-meteorite impact and piingement by various atomic and molecular
species. Thus one is faced with the difficult task of assessing the degree to which such an environment
will affect material prope-ties over a prolonged period.
By necessity, one must either transport the materials in :1estion to the sl .ce environment and
retrieve them for examination (such as is pli ned in tl e NASA Long Duration Exposure Facility for exanmle)
or test these materials in a simulated spac-i o"'o nt, Although a complete ground simalation is
virtutally impossible, the combined effects of tc,-xerature and --acuum are quite readily obtained. Addi-
tional features such as ultraviolet radiation and ha.j,. energy particle and molecular beam bombardment
can also be attained.
IF one considers the effect of vacuum alone, loss of adsorbed and absorbed gases as well as subli-
mation ci evaporation of the moxe volatile constituents of the material can occur. This is particularly
evident for ran orgunic/polymeric materials which are used as matrix or bonding agents in composite
materils, as illustrated in Fig. 1. Using a mass spectrometer capable of resolving species in the range
dc 13 ~ 80 AM4 and partial pressures of the order of 10"10 torr, it was found that the dominant out-
gassing component was H2 0+. Furthgrmore it car bg observed that it took about two weeks for the chamber
vacuum pressure to drop from 10- torr to - 10- torr due to material outgassing.
1 Moreover, if these
same materials are then exposed to ambient conditions, re-absorption occurs, as shown in Fig. 2. The
excess in weight ab)ove the initial sample value (prior to vacuum exposure) is probably due to moisture
diffusioa into the composite, occupying molecular sites vacated by other lower mass number polymer mole-
cules.
In eddition to vacuum and temperature, the cumulative effects of radiation and thermal fatigue pose
serious questions regarding the long terr behaviour of polymer matrix compositep in space. Recourse to
special protective coatings may vell be necessary to prevent significant material degradation.
The assessment of property changes resulting from exposure in a simulated space environment is best
done with the samples remaining in the vacuum state. This is desirable to eliminate poqsible ground
handling effects and re-absorption of atmospheric moisture and contaminants, as previousl noted. Thus
f
applications.
special facilities and test fixtures are required to provide in-situ loading, as will be described later.
By necessi';y, one must obtain ambient pressure reference data which are also of value for 'non-space'
Figure 3 summarizes the environmental and material parameters currently being investigated. To
provide a c~zjarison for the space simulation results, a composite experimnt (Fig. 4) has also been
assembled for inclusion in a forthcoming NASA IDEF miesion. Since this satellite is scheduled to be
retrieved a'ter about 9 - 12 months exposure, it will be possible to compare property changes, as out-
lined in Fie. 5. The purpose of this report is to present results obtained from one phase of the program
involving ground-based simulation. Material parameters of interest include strength, stiffness, coeffi-
cient of thermal expansion (CTE) and dexping.
2-3
+2 166O + 16
3 2 2o + 1
where it has been assumed that certain material property !yum.,try conditions hold true and odd order terms
in q6 vanish. This requires the lamina to exhibit identit ,-strength for both positive and negative shear
(relative to the principal axes). The principal strength ",.nnor components are defined by
1i i I
F F,= I F,2 = Y Y 6 S S'
1
Z
F 1 1 (2)
11Z XX F22 F66 S--T
and the interaction terms (Fl2, F112, FP2I, F166, F266) are a..cribed in Refs. I and 3.
Although the quadratic form of Eq. (1) is most often eriployed, one only has to examine the case of
biaxial loading to realize how deficient it is. For example, one can consider a symmetric balanced
laminate of a construction and calculate the failure condirion for varying biaxial tensile stress ratios
defined by Nx/Ny with Nxy = 0. Solutions were obtained in Ref. 2 for a graphite/epoxy material using both
the quadratic and cubic formulations. If one plots the maximum strength (OYLMAX) and corresponding opti-
mum fiber angle '6o' for each value of the biaxial stress ratio, substantial differences between the two
failure model predictions can be seen in Fig. 6. Note that the stress axis (GYLXAX) has been nondimen-
sionalized by the uniaxial tensile strength measured in the fiber direction, X m 185.63 KSI for SP-288
T-300 (Ref. 1). One interesting feature which emerges is that not only does the quadratic criterion
underestimate the strength, but the optimum fiber angle can be off as much as 100, although generally
the difference is less. This fact suggests that the quadratic equation cannot be used reliably in
strength design for selecting an optimum laminate configuration. Note that when a = 0.5, corresponding
to internal pressure loading for example, both formulations give the same angle. Figure 7 presents a
comparison of both failure models with experiments for this particular load case. It is readily apparent
that the cubic equation provides the best correlation with test data. Let us now demonstrate the effect
of a lamina failure criterion on the overall strength of a composite structure laminated in such a way
that first ply failure does not constitute total failure. Clearly, once a ply (or set of plies) has
suffered partial or total failure, a change in the stress/strain response should be evident. Referring
to the diagram of Fig. 8, several failure modes are possible: (i) matrix crazing or fracture with fiber
reinforcement intact; (ii) interlamina shear separation from adjacent plies; (iii) fiber/matrix fracture.
Consequently, after initial "failure", assumptions have to be made regard.-.: the effect of the "failed"
ply (or plies) on the stiffness of the remaining structure. In Ref. 2, internal pressure tests were
carried out on circular cylindrical graphite/epoxy tubes of ( el, 02)s sysetric construction. The
results from one particular experiment for an 8 ply ( 300, 55% tube are shown in Fig. 9. Multi-mode
F failures occurring at - 2 KSI and - 3 KSI agree rather well with the cubic model predictions. Note that
for this case the quadratic equation gives an overall ultimate strength of only - 2 KSI.
Having developed a failure analysis, it now remains to investigate the effect of various environ-
mental factors on the strength and stiffness parameters. For completeness, ene should include the four
V* lamina orthobropic elastic 'constants' (Ell, E22l vI2 and G12) as well as the five principal strength
terms (X, X', Y, Y', S). In general, knowledge of theas. parameters is sufficient tocalculate all of the
I tensor polynomial strength coefficients (see Eq. (1)] using the methods outlined in Ref. 1, and all of
the lamina stresses, assuming linear elastic behaviour. However, up to this point, environmental data
V| have only been obtained for E22 and Y, although one might expect that El, X and X' may not change appre-
ciably aftway.
2.2 Effect of Post-Cure Time and Cool-Down Rate
_J Prior to studying environmental effects on material strength and stiffness, it was decided to inves-
"tigatethe influence of manufacturing variables. Tests were carr~ed out on 4 ply 90" tubes for varying
autoclave post-cure times and cool-down rates.
In the first phavn involving post-cure times ranging from zero (i.e., no additional post-cure beyond
manufacturer's specift..ttons) to 24 hours, a glass/epoxy prepreg was used (3M SP 1002). It was found
Y that no significant chargeq in E22 occurred, although some ambient strength increaaas (Y) were obtained,
x' as shown in Fig. 10 in tL.,".s of F2 and F22.
- Anather interesting parameter from a mnufactu,ing point of view is the rate at which the cceposite
structure is cooled after the required cure cycle time in the autoclave. Varying rates of cool-down were
I achieved and recorded by thermocouples mounted on graphite/epoxy specimens. Typical rates of temperature
reduction are shown in Fig. 3 where it is indicated how the values of (dT/dt) ere estimatd. Again it
""t17 was observed that E22 di( not change significantly, but sew minor reduction in ambient strength (Y) with
increasing d/dt was found, as shown in Fig. 12.
2-4
Stiffness and strength data were obtained at ambient pressure as a function of temperature for a
graphite/epoxy material and are plotted in Figs. 13 and 14. The scatter bars indicate the range of maxi-
mum deviations found when three or more tests were conducted at a given temperature. It should be noted
that the ambient data were obtained from specimens stored under "normal" laboratory conditions, with no
regard for moisture content.
Based on the ambient pressure strength results, it was possible to calculate the thermal variation
in the graphite/epoxy strength tensor coefficients F2 , F2 2 as shown in Figs. 15 and 16, respectively.
From a design viewpoint, although this parameter is the "lowest" of the strength tensor components, its
effect on the ultimate strength of any laminate configuration can be quite severe -.Spending of course on
the nature of the loading.
As stated earlier, in-situ vacuum testing of polymer matrix composites is regarded as a necessity.
Thus space simulator facilities such as that shown in Fig. 17 have been developed at UTIAS capable of in-
situ loading to determine the stiffness and strength parameters. The vacuum system is composed of several
major components for pumping, valving, pressure measurement, baking, trapping and piping, a complete des-
cription of which can be found in Ref. 4. The stainless steel test chamber is cylindrical in shape,
measuring 51 cm in diameter x 117 cm in length. Combined action of rgughing and oil diffusion pumps with
I a liquid nitrogen cryogenic trap above the diffusion pump permits lO torr to be achieved in about one
hour. Of particular interest is the mechanical loading fixture attached to the chamber door (Fig. 18)
which can supply in-sit." uniaxial or torsional loading of specimens. Universal gimbals ensure alignment
of the load which is transmitted via hydraulic pistons acting through flexible stainless steel bellows.
It is fairly certain that for most fiber reinforced polymer matrix materials, the existence of any
thermal-vacuum effect should be most pronounced in terms of the matrix response. Consequently, test con-
figurations were selected consisting of graphite/epoxy 900 layups (i.e., circumferential fiber orienta-
tion) using thin-walled tubes (0.38 mm - 0.58 mm thick) manufactured in-house with a belt-wrapper appar-
atus. Using bonded foil strain gauges, the variation in E22 was measured under vacuum conditions for
varying temperatures (Fig. 13). If one now compares the mean results, it would appear that an increase in
stiffness occurs in the vacuum state. Because of the limited test data for T > 200 0F, it is premature to
speculate on whether the large differences in E22 exist or not. However, it is well known that moisture
content within the epoxy matrix can result in drastic strength reduction as one approaches he glass
transition temperature for the resin system. Since no precautions were taken with the "ambient" specimens
to limit or reduce their moisture absorpt'*on, there may well be a significant effect in terms of E2 2 as
indicated. Confirmation of this behaviour must await tests which will be performed on specimens currently
being studied in another facility.
Ultimate tensile strengths were also determined for the same specimens. A comparison between ambient
pressure and vacuum strength data for varying temperatures is given in Fig. 14, where the results have
been nondimensionalized by the ambient pressure, room temperature strength (YRT). Based on the mean
values shown, it would appear that an increase in the matrix strength parameter occurs in vacuum, although
substantial scatter in the test data is evident. Deskite this scatter, three anomalous failure loads were
obtained in the vacuum environment tha: fell well below the "scatter" rarge. From a design point of view,
ttie rether limited results available indicate a 20% probaoility of the occurrence of such low loads within
a grou.p of test specimens fabricated at the same time, fxom the same material batch under highly controlled
conditlors. Since this anomalous behaviour has never been observed in our multitudinous ambient pressure
tests conducted over the past five years using various batches of materials, it means that some random
"flaw" characteristic might be present whose effect is amplified under vacuum conditions, however, eluci-
dation of this effe^t should be forthcoming based on the number of specimens presently residing in the
long term thermal-',acuum facility. They should provide statistically valid data when strength tests are
carried out in the near future.
AI One of the main design considerations associated with the use of composites is the *
nojaibility of
achieving dimensional stability over a wide temperature range. For example, materials such as graphite/
epoxy and Kevlar /epoxy provide tht ewportunity to design, structural components with essentially a zero
value of CTE in prescribed directions. ZConsequently, an extensive program was undertaken to obtain CTE
i data for three materials: graphite/epoxy (3M sF288 T300), Kevlar /epoxy (3M SP306. FRD-49-Iii) and
and bo--on/epoxy (3M SF290). UT, to present, emphasis has been placed on determinf: g the variation in CTE
with "number of thermal cycles under both ambient pressure and vacuum conditions. It is of value at
this stage to present the theoretical CTE equation (refer toAppcndix A for the derivation) for a given
composite laminate. It should be emphasized that the structural CTE matrix [a) is dependent on the prin-
cipal lamina stiffness and CTE parameters as well as the particular filament orientation (e). Thus, for
the special case of i e symmetric laninates, the following equation can be used:
n
[ [Ai]I
i] *a= [ k hk-l)
k=l
wiLere
. [(q - and[1
10!,
0
IIL
A_1~
F 3.1
_2-5
All thermal response data were obtained using bonded foil strain gauges (type EA-06-250-BBl20, Micro-
Measurements, USA) after extensive calibration under ambient pressure and vacuum conditions. The apparent
response of a sorain gauged structure or material sample
reading one obtains from a strain indicator) is given by undergoing a temperature change (,M) (i.e., the
EA = "gg6 (4)
wfhere 'W' and 'g' refer to the 'material' and strain 'gauge', respectively. If the material and gauge
strains vary lneearly over the temperature range of interest, then we can re-write Eq. (4) as
-A
O CA+(X (5)
where amn and ag denote the material and strain gauge coefficients of thermal expansion, respectively. The
determination of an average value for ag at ambient pressure was made by bonding three gauges on separate
metal tubes: steel, copper and aluminum., subjecting each sample to uniform heating and measuring the axial
displacement as a function of temperature. This vas done in a small oven (containing a glass window)
mounted in a travelling microscope. At the same time, the apparent strains were recorded to correlate
with the actual strains at different temperature increments. After many repeat tests, a least squares
fit yielded an average value of Cag= 6.93 x 10-6 in/in/*F (Fig. 19). The next step involved calibrating
i the strain gauge-metal tube system in vacuum. As shown in Fig. 20, the calibration for ag shifted du to
outgassing and the final value used throughout the remainder of this report was taken as - 6.55 x 10-9
i (in/in/OF) under vacuum conditions.
Most of the subsequent test data are based on continuous thermal cycling (75*F < T < 200*F) at 10-7
~ 10-8 torr over a period exceeding 6 months using the facility shown in Fig. 21. T-his Titra high vacuum
chamber has a working volume of approximately 66 cm diameter x 76 cm in height. Employing a roughing
pump, two Zeolite adsorption pumps and an ion getter pump yields a vacuum close to 10- torr in a fully
loaded stata. Currently, 158 composite specimens (120 flats + 38 tubes) are mounted in a tray structure
slung below the top chamber closure plate (see Fig. 22). Of this number, 33 were strain gauged for moni-
toring thermal response of the materials. Both levels of the tray contain 3 thermocouples each to record
temperature. Note that throughout this experiment, a calibrated steel tube containing the same type of
bonded sti ain gauge was inside the facility undergoing the same thermal-vacuum cycling as the composite
samples. It was found that no significant deviation in the measuring system occurred, thus providing the
necessary confidence for interpreting the composite materials' response.
Figures 23, 24 and 25 present the CTE data obtained on 4 ply o symmetric balanced laminates as a
function of the nimber of thermal cycles for varying vacuum exposure times. The three materials investi-
gated include Kevlar/epoxy, graphite/epoxy and boron/epoxy
It is quite apparent that substantial changes occur in CTE due to the combined effects of vacuum and
thermal cycling. Since the development and accumulation of mtc rocracks together with bulk material out-
gassing probably account for this behaviour, it is reasonable to expect that the 'drift' in CT'E's should
level off with time. Clearly knowledge of the asymptotic values is required for design purposes. Inde-
L pendent measurements on a 43* graphite/epoxy tube confirm the vacuum effect change (see Fig. 23) after
only two thermal cycles for 48 hrs at - 10-7 torr.
""o assess test data consistency, it is possible to calculate the theoretical variation in CTE [Eq.
(3)3 with fiber angle based on knowing a 1 (ma 0o), a2 (Piag 0o) and the four elastic orthotropic constants,
El: E, G12, lv12. Tl.ui was done for azbient pressure and vacuum conditions, the latter corresponding
to 1E7 dayj at I0-7 ~ I0- 8 tort after 31 thermal cycles. Figures 26 and 27 provide comparisons of the
experinental results with predictions and it can readily be seen that excellent correlation was achieved.
Ncte that for the graphite/epoxy vacuum tests, only one set of data at 43* (other than al and a 2 ) was
available for comparative purposes. One other factor that should be considered in these calculations is
the variation in orthotropic elastic constants with temperature and vacuum. However, at present, we
.i "have only
10o-6 been
- 10-7 tortable to accumulate
up to graphite/epoxy
- 214 hrs (see Fig. 13). test results on the hsvlea
Noeh-suig change in E22 with temperature
2000F, at
curves were
fdesign constructed to compare with the room temperature (750F) modulus analysis, as shown in Fig. 26.
viewpoint, no significant effect n-"'rs at the zero CTE crossing.
From a
* REFERENCES
1. R. C. Tennyson, D. MacDonald and A. P. Nanyaro, "Evaluation 3f the Tensor Polynomial Failuze Criterion
for Composite Materials", J. Composite Materials, Vol. 12, 1978.
"N 2. R. C, Tennyson, A. P. Nanyaro and G. E. Wharram, "Application of the Cubic Polynomirl Strength Iri-
LI ili 3.
terion to the Failure Analysis of Composite Materials", to be published, J. Comnpositr Materials, 1980.
Materials,
S. W. Tsai Vol.
and E.5, M. Wu,
1971. "A General Theory of 3trength for AnIsotropic Materials", J. Composite
El- i 5
2-6
ACKNOWLEDGEMENT
The author wishes to acknowledge the financial support for various phases of this program received
from the National Research Council of Canada (Grant No. A-2783), the U.S. Nationel Aeronautics and Space
Administration (NSG-74o9) and the U.S. Air Force Office of Scientific Research (AFOSR-78-3694). The NASA
work is concerned with the composite strength studies whereas the AFOSR project involves space simulation
effects on CTE. Research associates and assistants participating in the composites program include: Prof.
J. S. Hansen, Mr. J. Catalano, Mr. G. Mabson, Mr. D. Morison, Mr. B. Uffen and Mr. G. Wharram.
IQ]11 Q22 0]
ni2 2 ]
._2
Substituting for the k-th lamina in-plane strains in terms of the laminate mid-plane strains and
curvatures, Eq. (A-l) becomes:
lxi I 'CO I aA
[Yy J =[ jlk o Ztijlk Q(A.2)
cyT Ky~ O
k k k
where Z is the + distance from the laminate mid-plane to the k-th lamina.
One can now integrate through the total laminate thickness to obtain the stress and moment-resultant
matrix [N/M), assuming comatibility of strain deformation between laminae and 6M = a constant. This
yields the following equations for IN/MN = EO/0]:
KT T 2 bo
Solving these equations for the mid-plane thermal strains and curvatures gives:
[Bij][A',j]li[B]}l { Z[][Qk a- ]( 2 2
K[11K]T
= { Di,'-]
k~lIk 0'2J
Sk=l
n r
1 0"
2-7
where
=
IK
[ K]
T LxY
and
1)T 1EY
Note that for the special case of a symmetric laminate,
[Bij3 = 0 and the above equations reduce to:
[K [D-ITDi-
J.l) I a (h- 0
I
{
[ -- IJ
y eyo/AT
one then obtains
S Eif1Z?[~ [ai
t k=l0
lipi-lQk
rja a [Aj[ (h
2 -
111
,
141
VI
2-8
S0.01;
I. K "s K
o QIO
8
100 x10
f0
-0
l-
01
w
0
(T*prtue~100F
I 1
10 30501020 5010 00
~2[
2-9
DEIIINSAEEVRNMNt"AIERHO9T
_ _ _ _OU GEO__
INPLMRMTI
EFET AEIL/CMOIE
MAEILSLCIN-ENOCDEOXE.TEMPATC,1LIIE
121:. NUBER
TubEsA 4.5U
63HNIA
F TET SECTIMNS
TH0.2A CYCs
-m
PHSCLi
45 FlasDIATION2.
MEHNIA LODIG c
j Fig.4
.E
UTIAS composi~~~~~~~FRATe
xeietfrteNS
THRALCN
OIGAS CM
ogdrto xoURE fct
ELECRICA
-aMPRO BETEE FLGTTS N PC IUAINDT
1'f
2-10
SPEIME
-Pg.5 es roram
LDEF
INSAS)
SI.I
TSTTnSACiACtCEC
il SII EPT
UTA
BIXATRESST AI ~'"
ISTORAGE
ItIA
tC-j TES
-D and L-E SPCMN RETIVL EXW
VI -M.~
2-11
2800-
2400-
0
0
0.In 0
2000-
IT U o Experiment
1600 X,&Predicted
S1200 -Cubic
~
uadratic,
Niff,800
5! i400-
1$ Load
Lorit Under
L:i.alcainge
rcdro Stheload-tanrsosn
ultimterstreingt fa lmnt
'4rnv
2-12
14
2.2 Wbwic First Failure.Il
Qujad. Ult Failure -2u i r --- Quad. Ult.Failure
Quad. U) 2060 KSI
14 -J
z
z , Laminate
1(+30,-30, 55, -55,-55,55,-30,30)
.6
Theory
x,o Exp't
-4 -3 -2 -1 0 2 4 6 8 10 12 14 x 10- 3
x 10" AXIAL STRAIN Ex HOOP STRAIN Ey
Fig.9 Comparison of cubic and quadratic failure models with experiment (SP-288 T300 3M)
F2 (KSI)I1
.3
01
, 0
.O2 4 8 12 16 20 24
(HOURS)
F2 2
0 44 2 1 0 ?
(HOURS)
Fig. 10 The variation in strength tensor component with post-cure time (glass/epoxy)
~-
I'..
2-13
250-
WI
SW.Normal Oven Cool Down
dlO-00 40 50 60
It) CHRS)
TIME
10 0n Doors
opSen
!We Coolo
80O
o o 0 30
ro
<dt
F-F
62000Y=16 L +75
""Fure 12 s
2-14
6
30-- xO (PSI)
E2 Effect of The rmal Vacuum Environiment on
E22 Tensile Modulus of Graphite/Epoxy
25 Laminate in 0=900 Direction
Vacuum Exposure S 214 HRS
( 3M SP 288 T300)
1.5"
VACUUM
AMBIENT (o" -to
1 .5'
I! ROOM rEMP
S III I I I II
Figure 13
1.2
08
-3 7 torr)
(3xlO_
06
0 Ambient (A)
0.4
: i0 0
,
I
I Effect of Thermal-Vacuum Environment on Tensile Strength of
GraphitelEpoxy Laminates (8 =900 Configuration)
II Figure 14
-
2-15
F2 0779
5'T
Material: 3M Compofly
Type SP-288T300
02-
KSI
01I
Figure 15
F22 y2-
Q221
0016'
Material: 3M Company
(-'7 Type SP-288T300
SI 0012
-44 I 9 I
Figure 16
1
2-16 A
zt~v~~S ~ c~OF
iiR
1w
1500 -x JO-6
U),
a Steel
cca CopperA
S000 - 4* Aluminum
w a
. a
A Fit Yields
A Q~Cgu.3IO (IN/INI*F)
Figure 19
0
2
6
~1000 ;10Y
Ambient Pressure ;0e
(from Fig. 19) ."
j -Vacuum
6
0 (10- -10-7 torr)
3 500- , aga6.55xlO01,N/IN/OF)
C')
0
0 50 100 150 200
tTEMPERAIURE CHANGE (F)
Figure 20
IAR
-. 2-18
1i.2 Ta
oti n opst ape
2-19
30-90 ------------
75OF s T s 200OF
20
z I W.^O;
!N VACUUM
01
-0
-10-
0 6 12 18 24 30 36 42
N. THERM4L CYCLES IN VACUUM
GRAPHITE / EPOXY
6 "(31M SP 288 T300)
5 -x10-
AMBIENT VALUE
z 0 +
+'..I
43*I
t 43'
Figure 23
40 ~ .
60'
30-
3C
N0
l.0 6 12 18 24 30 36 42
./. No. THERMAL CYCLES IN VACUUM
"Figure24
2-20
BORON / EPOXY
30 X106 (3M SP 290)
S20
0N 90 75 0F S T' 200OF
z 10- )=DAYS IN VACUUM
z
SC *
00 I I I '
(167)
(91) (126 )
(5N) 6[
0 12 18 24 30 36 42
No. THERMAL CYCLES IN VACUUM
Figure 25
-6 GRAPHITE / EPOXY
40 -X 6 (3M SP 288 T 300)
30 ( 0 0-
I
Vacuum
No T/C = I(0-7-10-8
"20 , T im e in Vacuum 2 00 "Fvo 7
_z
167 days
z t r
lo m-bient
-S-
'C 2006F
2000F 750F
04 60 80
-I FIBER ANGLE e*
-10- 75"F
"i ' "Figure
26
rill
4t,!
2-21
60 - x 10-6
Vacuum
x
i~50-
No T/C 31 (10-7- 108 torr)
Time in Vacuum
;. 40-
40 167 days
4-30
h
8-
-
0
_z
"-20
I0
20 60 80
-9 _ 4 FIBER ANGLE 0e
KEVLAR/EPOXY
-10- (5M SP 306)
'. -20L
Figure 21
II I
I I
P
7 3-1
Summary
Graphite/epoxy mateiials are finding widening acceptance in the aircraft industry. Their
generally attractive strength properties, however, tend to degrade in hot and moist envi-
ronments. The high dependence of the strength loss on the moisture content demands an as-
sessment of the amount and distribution of absoibed moisture, particularly in the case of
light-weight sandwich panels. The reliability of such predictions hinges on the formula-
ticn of theoretical considerations, the accuracy of numericalprocesses, the definition of
material constants and a proper interpretation of the environmental conditions. The impact
of errors in these parameters on oredicted moisture contents, and comparisons of analyti-
cal forecasts with experimentally determined values, are the topic of this paper.
1. Introduction
Graphite/epoxy laminates tend to absorb moisture from The atmosphere, principally by dif-
fusion through the epoxy matrix. The bonding of the water molecules to the hydroxyl groups
of the epoxy polymers leads to a swelling of the matrix and to changes In the state of
stress
ature ofin the
the matrix,
laminates.
i.e.,The
it presence
softens of
the moisture
resin andalso lowersdc'rades
thereby the glass
the tran:sition temper-
matrix-dependent
laminate porperties at elevated temperatures. The severity of the degradation depends on
the amount and the distribution of the moisture which are continually affected by changing
temperature and humidity conditions. Drying of the moist laminates seems to restore the
laminate properties fully unless permanent damage to the matrix material is caused by ther-
mal shocks.
Apart from temperature and moisture, the properties of graphite/epoxy laminates depend, of
courseon the characteristics of the epoxy, the fibers and the fiber treatment, on ply
orientation, resi.i content, curing procedure, etc., so that a prediction of their actual
strengths or st.ffnesses under given conditions is difficult. An evaluation of a multitude
of test data fiom different sources[llfor T 300/934, T 300/5208 and AS/3501-5 materials,
however, leads to the interesting observation that the degradation of their compressive
strength under hot and moist conditions can be approximated, in general terms, as percent-
ages of their initial dry strength at room temperature. The data leading to the failure
surface depicted in Figure 1 were reasonably compatible in the sense that all test speci-
mens had comparable ply orientations and thicknesses were cured in accordance with the
supplier's reco~nmendations,cnd were tested in the form of sandwich beams under four-point
loading. These data indicate a uniform trend of several materials towards alarmingly re-
duced strength properties when exposed to moisture and heat.
The moisture problem is particularly sensitive i'nthe case of graphite/epoxy sandwich pan-
els with moisture-absorbent cores. Here, the structural safety at elevated temperatures
is threatened not only by the degradation of the faceshents but by internal pressures gen-
erated by the vaporization of moisture contained in the core material as well. Figure 2
shows calculated vapor pressure intensities developing inside a sandwich panel with Nomex
core after saturation at 40 % and 100 % relative humidity, and subsequent heating to 175 0 C.
Corresponding tests have shown that, under adverse conditions, such panels will explode
due to internal pressure alone.
Considering the critical nature of the moisture-induced degradation, it is obvious that
the design and the qualification of graphite/epoxy structures destined for operation at
high temperatures mandates a quantitative evaluation of these effects. Such an evaluation,
generally, requires a forecast of the worst expected moisture state by analytical means,
and a subsequent determination of the remaining margins of safety by test.
It is commonly agreed that the accuracy of analytical estimates depends on four factors: 3
0 The applicability of the classical theory of diffusion to fibrous composites;
The following discussion attempts to describe the significance and limitations of these
factors, and to assess the accuracy of moisture predictions based upon them.
I
,. composites is not obvious but has been repeatedly claimed by several investigators[2],[3],
for the one-dimensional flow of moisture through the thickness of laminates of large later-
al dimensions.
~
3-2
Idt
and
(5) am D a2M
~; (TaX
Recent experiments with T300/Code 69at the DFVLR question the supposition of negligible
concentration dependence. According ho F~gur? 3 liediffusion coefficients measured 9t
room temperature varied from 1.4x1O mm s- at 27.5 % relative humidity to 3.3x0- s
at 95 % relative humidity. Apart from this aspect, unresolved problems exist in regard to
the influence of ply orientation, fiber treatment, surface conditions, and others. On the
whole, however, it may be stipulated that the theoretical basis for the prediction of one-
dimensional moisture flow in graphite/epoxy laminates, while far from being exact, leads
to useful approximations of the moisture state.
(6) am 1 - am
The second condition requires compatibility of the moisture levels at the interface of tile
two materials. Definingas percentage of the laminate dry weight, as M the moisture
saturation at RH = 100 %, the moisture saturation at values of RH <18801 can be expressed
"in the form
4I,
3-3
The equality of (RH),., and (RH) at the interface has been deduced by the assumption of
a fictitious airspace between thl materials of infinitesimal width [31 but has yet to be
proven by a more rigorous argument.
Sandwich panels, in a general sense,ca.a be analyzed as multilayered slabs if the core ma-
terial is regarded as a homogeneous layet with uniformly distributed densities and diffu-
sion characteristics. In the case of honeycomb cores, this idealization is inadequate be-
cause the predominant portio.a of the moisture transfer occurs normal to the surfaces of
the core ribbons through the entrapped air in the cells. A more realistic treatment of
the problem is the coupling of the adhesive layers to both sides of the core by means of
an internal moisture belance[3]. A welcome by-product of this approach is the calculabili-
ty of the amount of moisture in the entrapped air space, and the corresponding vapor pres-
sure. Figure 6 shows the physical and mathematical models of a honeycomb-type core with
layers of adhesive on both sides. The difference of the fluxes during the time interval
At represents the weight of moisture absorbad by the core, W , and that of residing va-
por in the cavity, Wv. Introducing as w the weight and ag v tbe volume of the core per
unit area of the panel surface, then with Wc = w M10 0 % H and Wv = PsV/RT it can be shown
that
4. Material Properties
The prediction of moisture contents and distributicns in graphite/epoxy sandwich panels
depends critically on the correct identification of material properties which affect the
moisture response, i.e.,
o Maximum absorptivity at 100 % relative humidity, MI0 0 %;
o Maximum absorptivlty
in Equation (7); at relative humidities <100 %, MRH,
MI character'.zed
. IThe
from zero at RH = 0 % to M10 0 % at RH = 100 V.
tempepature dependence of the diffusion coefficients can be expresseg in the form D =
D exp (-) , where D and C are constants, and T is the temperature in Kelvin. Accepting
Ris rela~tionship as Valid, the values of diffusion coefficients, plotted on a lug D versus
onstants D and a limited
fromtaken
thatalso test, data
number of [4) the
-graph, ought to C can or.
fall be straight
evaluated. Figure
lines, so 8, from Reference displ~ays
8
diffusion c efficients experimentally determined by several investigators. The data scat-
ter is appreciable and the evaut igr 8, a levels will Rare accordingly. Figure 9
shows the sensitivity of the predicted moisture absorption at 21 9C with respect to dif-
ferent values of diffusion coefficients reported for T 300/934. and a comparison to actual
absorption values determined by weighing.
It follows from Fick's laws that absorption and desorption are 9overned by the same prin-
I_ _ _-_ _ _
Z;1'_4 -4
3-4
ciples and differ only with respect to their direction, i.e., the diffusion coefficients
should For graphite/epoxy laminates
this beis numerinally
for equal.
subs absrption and desorption
a, tial evidence exists that approximatell correct for temperatures up to
100 C, while contradictory test data were obtained for T 300/934, T 300/5208 and T 300/
F-263 in Reference [11] for td1nperatures above .OOC. ThS shape of drying curves observed
duiLng 24 hours of drying of saturated specimens at 120 C, is not consistent with the
standard diffusion laws. T~e problem is illustrated in Figure 10.At 120 C, a diffusion
coefficient of 600x1O-8 mm /sec fits only the first part of the drying curve reasonably -
well. Beyond that the drying curve corresponds more to a diffusion coefficient of 300x0O-8
mm2/sec.
This observagion is baffling but perhaps explainable by the fact that the drying tempera-
ture of 120 C lies above the glass transition temperature of a saturafed epoxy matrix,
and that different diffusion coefficients may associate with the crystalline and the amor-
phous phases of the epoxy. Figure 11 shows that good agreement between theory and experi-
mental data can be obtained with the assumption of a moving boundary between the two pha-
ses which, effectively, requires a numerical solution of Fick's equation in which the dif-
fusion coefficient D is a function of concentration.
7. Recommendations
The preceding arguments were stimulated by the senior author's experience with graphite/
epoxy sandwich components of the Space Lauttle Orbiter. There, the tracking of the moistu-
S re flow led to useful approximations of the residual material properties because the ini-
tial condition was known in the sense that, after six months exposure to the rather con-
stant climate prevailing at Palmdale, California, moisture equilibrium at 30 % relative
-. humidity could be assumed. The subsequent conditions were also well defined by a precise
f time-line in controlled environments.
In the majority of aerospace systems, however, this situation does not exist. Especially
with respect to aircraft structures designed for long life-time in variable climates, cal-
Et culated predictions are hazardous. Apart from other imponderables the uncontrollable cli-
3-5
matic fluctuations encountered in regular service defy any sensible analytical attempt.
The thought lies near to circumvent the entire issue of moisture degradation by covering
the exposed surfaces of the sandwich panels with a sealant. Since organicbecause substances can-
of their
not De expected to provide an effective seal over extended time periods Their attach-
own inherent diffusivity, the only feasible alternative are thin metal foils. the likeli-
nent without tears or wrinkles is difficult especially on large surfaces,of and moisture through
hood of externally caused damage must be accepted. The gradual entrance
foil and the ensuing lateral distribution over extended time spans
small openings in the cannot es-
is dangerous. It is possible that in hot environments the accumulated moisture separation of the
cape quickly enough and that the developing vapor pressure will cause aseparation under ad-
foil from the sandwich facesheet. Figure 15 is a vivid example of foil cured sand-
mittedly severe conditions. The thin aluminum foil was bonded to the previously Nomex
wich panel by a hot-temperature-curing adhesive. The panel containinganda perforatedthen exposed to
coreowas subsequently saturated in a 95 % relative humidity chamber
of the resulting blisters is solely due to the expanding vapor from
175 C. The formation and questions the wis-
the moisture contained in the adhesive and the sandwich facesheet,
dom of applying metal foil seals.
The currently emerging trend deals with the moisture problem by anticipating complete during its
saturation at the worst average humidity level to which the aircraft is exposed recognition that
life-time. This maximum average must be defined judiciously and with the
will not be uniform in the thickness direction of the laminate
the moisture distribution This
but will exhibit gradients commensurate with the fluctuations of the environment. dif-
point is illustrated in Figure 16 which shows, for the same total moisture content,
ferent distributions inside the laminate. It ought to bf expected that such variations of graph-
from a uniform distribution significantly affect the matrix-dependent properties seem
ite/epoxy laminates. Recent tests at the DFVLR indicate, however, that such effectsthe ply
to be quite moderate. In the case of laminates made from T300/914 C material with
differ-
sequence [45-18 and subjected to tension loading at 130 C, no substantial strength 1 %
ences were noted in test specimens with similar moisture contents of approximately encouraging al-
but different distributions. The test results summarized in Figure 17 arereplicates only,
though it must be recognized that the test points are averages of three of this
and that no conclusions can be drawn from such a limited data base. The importance conserva-
issue, whose as yet uncertain consequences force the adoption of perhaps unduly
tive safety factors, suggests further testing of different ply sequences under various
environmental and mechanical loading conditions.
8. References
[1] Collected heat data (1974-76) from Rockwell International (Los Angeles and Tulsa),
McDonnel-Douglas (St. Louis), Lockheed (Sunnyvale) and ITTRI.
[2] C.H. Shen and G.S. Springer, "Effects of Moisture and Temperature on the Tensile
Strength of Composite Materials", J. Composite Materials, Vol. 11 (1977).
K. [3] D.J. Zigrang and H.W. Bergmann, "The Response of the Space Shuttle Orbiter Graphite/
"A Epoxy Sandwich Panels to Exposure to Moisture and Heat", Proceedings, 11th Congress
ICAS, September 1978, Lisbon.
[4) A.C. Loos and G.S. Springer, "Moisture Absorption of Graphite/Epoxy Composites
Immersed in Liquids and in Humid Air", J. Composite Materials, Vol. 13(1979).
"Moisture Diffusion in Graphite/Epoxy Laminates:
"[5] R.E. Bohlmann and E.A. Derby, Presented
Y Experimental and Predicted", at the 18th Structures, Structural Dynamics
and Materials Conference, and at the Conference on Aircraft Composites: The Emerging
Methodology for Structural Assurance, San Diego, Calif., March 1977, AIAA Technical
Papers. Vol. A (A77-25726, 10 39), pp. 219-226 (1977).
Composites",
[6] R.De Iasi. and J.B. Whiteside, "Effect of Moisture on Epoxy Resins andJ.R. Vinson, Ed.,
Advanced Composite Materials - Environmental Effects, ASTM STP 658,
American Society for Testing and Materials, 1978, pp. 2-20.
17] J.M. Whitney and C.E. Browning, "Some Anomalies Associated with Moisture Diffusion in
Epoxy Matrix Composite Materials", Advanced Composite Materials-Environmental Effects,
1978,
ASTM STP 658, J.R. Vinson, Ed., Americar Society for Testing and Materials,
(~ pp. 43-60.
18) C.D. Shirrell, "Diffusion of Water Vapor in Graphite/Epoxy Composites", Advanced
Composite Materials - Environmental Effects, ASTM STP 658, J.R. Vinson, Ed., American
"Society for Testing and Materials, 1978, pp. 21-42.
1[9E.L. McKague, Jr., J.E. Halkias and J.D. Reynolds, "Moisture in Composites: The Effect
pp. 2-9.
of Supersonic Service on Diffusion", J. Composite Materials, Vol. 9 (1975),
i!;
-_
-4 -
F'7
3-6
,!Residual Compressive
Strength (%)
100
~p \,"
93
Fig. 1 DEGRADATION OF
i! COMPRESSIVE STRENGTH
;ii7.2 _ .
-.. 180
,i: c coreatsaturation0 /
Pressure
100%
5.4 -_ _ _ 5 I--
7. -
8
0 '-60
0 500 1000 1500 2000
S, Elapsed Times t (sec)
Fig. 2 INTERNAL PRESSURE IN MOIST
:I SANDWICH PANEL
Material: T300/Code 69
Ply Sequence: 10' 45 .0 1,,-
/ exprimnta
ratureConcentration De-
C0) 4. Fig. 4
experimentalConceutrationaDe
Normalized
Epsr Time,Af
lb 5b_____6 10 00 004dO50
ADH CORE AOH ADH ADH , ,
~~CORE OID,
SVOUM Fig. 6
> . .Core Ra 'resentation
S~
VOD
VOLUME
Temperature, T (CC)
140 100 50 20
Ii ., I II
; i
E 0O\ Fig. 8
10 Diffusivity as Function of
>4, 0 Temperature for T3OO/1034
,.ji
3-9
ID :2.0 -
(n 1.5
/C~alculations
D = 310-8
with
l10 15 20 25 30
Exposure Time (Days)
20~
I IMaterial T300/934
__Thickness :1mm
Calclatons ith Resin Content: 28%
1. -D =275 M[95%) :1,74%
* 300
D Drying Temp. :1201C Fg1
I DD=375 Experimental and Calcu-
0 D 0 m~sC) 2S _______lated Drying curves
j _05
e Experimental Data
Material : T00/934
Thickness :1mmn
Resin Content: 28%
D~. M19 5 %)' :1.74%
-T Dryirng Temp. :1250C Fig. 11
______ ___ExpeimetalAdjusted Drying Curve
e Exeimna Data
V,"
) -ClcultedResponse with
0 cytiie 2001 sec
_0_5
0 5 10 15 20 25 30
Exposure Time (Hours)
3-1O
GRAPHITE/EPOXY
(32% RESIN BY WT) 3,1 17 320 6500 2,0 0.268
(52% FIBER BY VOL)
TEST ANALYSIS
PANEL NO 1 F/S 1.6% 1.5%
DRY PANEL. EXPOSED ADH 2,3% 17%
408 HOURS TO 95% RH
AT 60 0C CORE 3.0% 4,1%
10.0 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _= ___ _ _ _
7.5-1
( 5.0. -
Rel. Humidity :95%
- t---Temperature: 600C
o Experimental Data
Fig. 15
SANDWICH PANEL
WITH BLISTERED
j~h ALUMINIUM FOIL
Material :T300I9i1.C
:1Thickness .2mm
E 0.Resin Content: 26%4
E~ 180. Ply Sequence :f45. 451,
- 6 Test Temp. : 1301C
fI ~140O
~120-
Reference Fia.16
by
Renato Barboni
Mario Marchetti
Isidoro Peroni
Istituto di Tecnologia Aerospaziale
dell'Universiti di Ron-
Via Eudossiana, 16 - 00184 Roma - Italy
SUMMARY
Exnerimental and analytical results are presented for the mechanical and thermal behavior of carbon fibre pultrusion and
laminate. Particular attention is paid to the temperature effects on the static mechanical cheracteristics and their non hookean
behavior for the range 300 + 450 K.
Dynamic tests in crder to find the internal damping are also carried out. The linear coefficients of thermal expansion, a,
are reported for the temperature range 150 + 400 K.
It is shown that calculated values of thermal expansion are in close agreement with those measured.
II
1. INTRODUCTION
As carbon fiber composites are still, by engineering standa.ds, a relatively new material, there are serious problems in the
knowledge of their characteristics primarly the variability between apparently identical composites and everybody operating in
the field of materials should endeavor to exploit their properties as much as possible.
The paper deals with the composite material consisting of continuous carbon fibers ar.anged with the long fiber axes aligned
in resin matrix (CFRP).
Classical laminating theory and different models [1], [2],[31 has been applied for predicting the elastic and thermal pro-
perties of CFRP from properties of the loyers and of the constituents too.Nevertheless theprevision of the real values is
V complicated by many factors such as the variation of carbon fibers in the cross sectional [4] and the variability of th,
constituent material properties to environmental effects [5], [6], [7].
SAt same time the experinmental characterization of CFRI' is generally complicated to macroscopic anisotropic
and various coupling effects.
Only a comparison of the results predicted by theoretical approaches and experimental tests can give a
better understanding of the CFRP behaviour.
In Aerospace Technology Institute of Rome are available computer programmes to predict the mechanical properties
A (Young's moduli, Poisson's ratios, shear modulus, ultimate strength) and thermal properties (thermal expansioit coefficient and
thermal conductivity) reiating to the lamina and the laminate too from the raw material properties of matrix and fiber [8].
In order to check the theoretical results on the mechanical properties at room temperature an experimental research was
; carried out last year [9].
i "i The main
Properties CFRP.of the present paper is to reveal certain temperature dependent behavior of mechanical and thermal
of thepurpose
The necessity of this research derived by considering satellites which are at present in early development stage and the
italian industry SELENIA, have examined the opportunity to design some of their structural member in CFRP.
As a matter of fact the high thermal stability and specific stiffness make CFRP an obvious material choice for antenna
j
'; dishes and struts [10].
An analytical and experimeital investigation on temperature effects of pultrusion and lay-up laminates is carried out.
It is shown that calculated values are in close agreement with those measured.
4-2
2. SPECIMENS
For the nresent purpose two fundamental set of composite type was utilized:
a) Pultrusion compounds: the specimens source was bars of rectangular cross-section prepared by Courtaulds with fibers
first passed through an impregnating bath of matrix material, and then pulled through a heated die of the rectangular cross.
section where the composite is cured.
Micrograplhs of transverse and longitudinal sections of typical bar is given in fig. 1,, which show uniforme fiber distribu.
tion, absence of voids or cracks and excellent fiber alignement.
The epoxy resin is designed by the manufacturer Epikote 828/MNA/K61B and the fibers GRAFIL A-S.
The values of the physical and mechanical properties are reported in Table I.
3
Density (g/cm ) 1.8 1.2
Ultimate Tensile
Strength Stegh(~)2.30
(GPa) - 2.64 0,07
b) Laminate: the specimens were cut out of two sets of source plates made in Rome with prepregs delivered from two
different suppliers:
B - Courtaulds: the preimpregnated fibre was prepared containing GRAFIL A-S fibre in Shell DX 210 epoxy.
C - Carbjform: the prepregs with code 87 resin system and with GY - 70 carbon fibre was utilized.
The plates were fabricated by the following autoclave moulding process: a charge of different layer orientations was
placed in an autoclave which was pressurized to 30t mm of mercury, the composite panel was heated to 120 + 125 *C with
1 rate of 2*C/min and a pressure of 600-. 650 K Pa was applied, the heating was continued to 170 3 C; this temperature was
holded for 60 1 5 nins and the autoclave was cooled to 100: 5 OC; at least the pressure was released and the laminate re-
moved.
Electronic microscope enlargements of transverse and oblique sections of one obtained laminate are shown in Fig. 2.
The elastic and thermal constants for the cast resin and iber are listed in Table It.
J" ,4j
Z
4.3
Fiber Volume - 60 40 40
Fraction (%)
3
Density (g/cm ) 1.87-
Ultimate Tensile
Strength (GPa) 2.35
Elastic Modulus
3.5 3.5
(GPa) (a',345 .
-I1 (L)
Thermal expansion
coefficient (10-6/K) ~ 65 ~ 65
+ 17 (T)
3. THEORETICAL PREVISION
Analytical studies of the mechanical behavior of a layer via micromechanical formulation were developed. This calculation
permitted us to perform subsequent numerical analyses in which the actual heterogeneous material was conceptually represent-
ed as a homogeneous, anisotrcpic solid via classical laminate theory.
In a preceding paper 18, a description of the background to micromechanics and macromechanics was followed by the
computer program in which particular attention was paid to the influence of fibre type and orientation. For the present
purpose in order to asses the influence of the temperature the programme was implemented to take into account the varia-
bility of the constituent materials properties.
The properties of the carbon fibres were ass imed to be unaffected by temperature and those of the matrix only were
assumed to be temperature dependent, over the range considered in the analytical prevision.
Numerical results of the computer program will be presented later in comparison with experimental values.
L 4. STATIC PROPERTIES
The tensile test data reported in this paper was dete-mined using the testing machine shown in fig. 3, equipped with a
specially constructed electric oven, in the temperature range of 300+ 400 K.
The dimenrions of the heating cylinder werv such that all the specimens were completely inside the box. Termocouples
placed inside the oven showed variations in temperature of the order of 0,5* C from top to bottom of the specimen.
Four fundamental group of specimens were tested A/a, Bl/a, BI/b, Cl/a, C2/a, summarized in Table Il.
7, Dimension of test specimens are shown in fig. 4, with effective variation within 1 percent. All the samples were the
straight sided specimcns but only the laminates with end aluminium tabs.
4-4
Each specimen was instrumented with two biaxial strain-gaugeson opposite faces and the tensile stress-strain curves were
plotted.
Procedures for temperature testing consisted of mounting the specin'en in the load frame with end c9nnections of serrated
jaw type, [11], a few minutes (5 15) to raise the temperature to the desired level, an approximately equal number of
minutes temperature soak to uniformly distribute the temperature, and ther, s.nce composites may behave visoo-elastically the
load was applied linearly to rich the maximum value in 20- 30 seconds.
Young's modulus and Poisson's ratio wcrc deteiminied by tensile tests which were carried out to approximate!y 20% of the
tensile strength.
It has been observed that,with regard to each sample group, the specimens tested have been show, with ver) small scatter
(< 8%) a non linear stress-strain relation 112].
This non hookean behavior, that increases with temperature, can be one of the most important reason for the remarkable
scatter found in the results of tensile tests from various reserchers.
In order to give a reference point the Young's modulus a'nd Poisson's ratio reported in this paper as function of tem-
perat ;Se was obtained at o = 0.1 GPa.
k more detailed analysis of the results is given in the followings:
Type A/a : For pultrusion compound the temperature effect on Young's modulus can be seen neglegible and the
theoretical prediction agrees very well with experimental results, fig. 5.
All tensile tests show a non linear stress- strain relation fig. 6, and the relation between Young's modulus and s~ress is
plotted in fig. 7.
Type B The value of Young's modulus as function of the measurement angles at room temperature is shown in
fig. 8 as a comparison between experimental and theoretical results.
The temperature effect was neglegible for tensile test on samples with measurement angle 0 = 90V but for sample% with
measurement angle 6 = 450 was significant as it is shown in fig. 9. The same consideration is valid for Poisson's ratio
fig. 10.
The variation of E with respect to stress for the samples of BI/b group is shown in fig. 11.
The Young modulus first increases with stress then decreases; this can be imputate to opposite effect of fiber and resin;
the effect of resin becomes predominant with increasing of stress and temperature.
The variation of Poisson's ratio vs. stress is given in fig. 12.
Type C All tensile I-sts show non linear stress-strain relation, fig. 13. With respect to this non-linear relation, the
relation beetween Young's modulus and stress was found to be linear as sho-.n in fig. 14 -15.
The values of Young's modulus vs. the temperature for samples of the groups Cl/a are plotted in fig. 16.
A most important effect of temperature on E results for samples of the group C2/a, where the lay-up orientation is
not parallel, is, shown in fig. 17.
The variation of Poisson's ratio vs. temperature for the group C2/a is given in fig. 18.
S. DNMCPROPERTIES
We present experimental results obtained from a series of free and "sinusoidally forced vibration tests of pultrusir', com-
pound in order both to verify the validity of the previous results and to find the internal damping [131.
Specimens were clamped at the ends and excited in first resonant flexural vibration by an electromagnetic shaker while
maasurements of frequency response were made and recorded. Since one of our objectives was to study the effect of the
Svacuum, specimen were tested in a small vacuum chamber, fig. 19, at various pressure in the range from room pressure to
I10-7 m'mHg.
Measured values of specimens resonart frequency (380 Hz) and specimen geometry were used to find the longitudinal
Young's modulus by utilizing Behrens correction [II].
The obtained value at room temperature was E = 109.7 G Pa; the scatttring results less then 5% with respect to data
shown in fig. 7.
The experimental values of the dimensionless damping ratios , as function of response amplitude at room condition are
given in fig. 20.
The importance of the vacuum is shown in fig. 21; the effect can be estimated as a reduction of 15% in the dimen-
sionless damping ratio.
- a . .
4-5
The thermal expansion coefficients were determined over a temperature range of - 150 + 450 K utilizing an electronic
dilatometer in which the measuring device reaction not exert a stress of more than 6.10-2 N/mm 2 on the specimen so that
the specimen is not intended.
The samples were cut from pultrusion bar so that the fiber axis was at 0 = 0, 150, 30Wand 450 with the edge of the
2
source bar. The final sample measured 25 mm in length and 5 x 5 ram in cross-sectional area.
Each specimen was heated at the approximate rate of 1.3 C/min in the dilatometer, fig. 19, from room temperature to
450 K and then cooled to initial value at same temperature rate. A small amount of hysteresis was observed between the
heating and cooling curves of the first cycles. Further heating followed very closely the cooling curve and the behavior be-
come essentially reversible. A minimum of four cycles without stoppage of work were recorded for each sample so that
reproducible expansion curves were obtained.
The some procedure was followed in the range - 150 370 K -..e iiy-.
.y esis; in those thermal cycles, temperature
higher then room temperature was reached to verify the continuity of the behavior below and above room temperature.
In general, three specimens were run for each determination of expansion coefficient and a variation in individual values
was less there 10%.
In order to compare the experimental values, the prediction of pultrusion thermoelastic properties from constituent pro-
perties has been carried out.
The material properties vs. temperature of the epoxy as the input data to the computer program are shown in fig. 20
in agreement with [14]. Unfortunately these properties concern only to the temperature range between 300-+ 450 K so that
the prevision not was carried out outside this range.
The experimental and theoretical data on the linear thermal expansion are summarized in fig. 21, for the four angle
orientations. In the temperature range subjected to inspection the linear thermal coefficient can be cons' -d approximatly
constant for angles 0 s! - 300 . 300.
As matter of fact the role of the resin becames predominant with the increase of the angle of the fiber orientation,
0, as might be expected [14],[15].
It can be observed that agreement between the experimental and the corresponding numerical results is 15ood.
Off- axis experimental thermal expansion coefficients as a function of angle with the fiber orientation are shown in fig.
22.
The measured values at different temperatures and those calculated are plotted in fig. 23; the experimental results agree
the numerical results very well.
7. CONCLUSION
It is hoped that this report has touch some of the most important problems connected with the exposed surface and
struts of satellite structures.
It is not the complete answer to all satellite structural problems but other aspects such as various environmental con-
dition, fatigue and impact resistance, implication of laboratory accelerated tests ... are discuss at this specialist meeting.
Grateful acknowledgement must be given to responsables of the AGARD Structures and Material Panel for the meeting
arranged to cover both theoretical and practical aspects of the composite materials behavior.
At the Institute of Aerospace Technology of Rome is already begin a very extensive research, supported by ESA, in
order to validate a general numerical program by an intensive program of experimental tests.
"More then 500 specimens will be tested and this is the right way to realize the advantage of the material and to in-
troduce it into satellites.
~I
t ;
4-6
REFERENCES
[1] G.P. SENDECKYJ: "Elastic Behavior of Composites". - Composite Materials, Vol. 2. Academic Press 1974.
12] C.W. BERT: "Analysis of Plates". - Composite Materials: Vol. 7. Academic Press 1975.
[31 E. REISSNER and Y. STAVSKY: "Bending and Stretching of Certain Types of Heterogeneous Aeolotropic Elastic
Plates". -- J. of A.M. September 1961 p. 402.
[4] HAWKER SIDDELEY DYNAMICS Lt - SPACE DIVISION: "Final Report for Phase I of a study on !he use of
Carbon Fibre Reinforced Plastics in Satellite Structures". TP 7427- 1973.
[5] C.H. SHEN and G.S. SPRINGER: "Environmental Effects on the Elastic Moduli of Composite Materials".
J.C.M. Vol. Il/ July 1977 pg. 250.
[6] C.Y. LUNDEMO and S.E. THOR: "Influence of Environmental Cycling on the Mechanical Properties of Composite
Materials". J.C.M. Vol. 11/July 1977 p. 276.
[7] C.H. SHEN and G.S. SPRINGER: "Effects of Moisture and Temperature on the Tensile Strength of Composite
Materials". - J.C.M. Vol. 11 I/ January 1977 pg. 2.
[8] R. BARBONI and I. PERONI: "Mechanical and Thermal Coefficients in Composite Materials". - Istituto Tecnoiogia
Aerospaziale - Q.C.E. n. 23 - Roma 1979.
[9] AA.VV.. "Studio di strutture a tecnologia avanzata e ad alta stabiliti per antenne di un satellite di telecomunicazio-
ni". - Research supported by CNR/SAS - GLISSCAS Final Report - Rome Oct 1979.
[I1] C.W. BERT:- "Experimental Characterization of Composites". - Composite Materials, Vol. 8 Academic Press 1975.
(12] W.H.M. Van DREUMEL and J.L.M. KAMP: "Non Hookeai Behaviour in the Fibre Direction of Carbon-Fibre
Composites and the Influence of Fibre Waviness on the Tensile Properties". - J.C.M. Vol. 11 October 1977, p. 461.
[13] P. SANTINI, A. CASTELLANI and A. NAPPI: "An Introduction to the Problem of Dynamic Structural Damping".
Agard - R-663- 1977.
[14] T. ISHKAWA, K. KOYAMA and S. KOBAYASHI: "Thermal Expansion Coefficients of Unidirectional Composites"
J.C.M., Vol. 12 April 1978, p. 153.
[15] B. YATES, M.J. OVERY, J.P. SARGENT, B.A. McCALLA, D.M. KINGSTON-LEE, LN. PHILIPS and K.F. ROGERS:
"The Thermal Expansion of Carbon Fibre-Reinforced Plastics". J.M.S. Vol. 13, p. 433, 1S78.
Ii
SACKNOWLEDGEMENTS
The authors wish to acknowledge the help of Mr. Siclari for his technical assistance in performing the experiments
and the Selenia Spa for the preparation of the specimens.
112
4-7
44
'13
4-8
A.;.
Hjr
Ii
It
150
Hi
PULTRUSION
-L5
I
"ayi
~
S100
Fig.4
1
E[GRPal
120.
100. EXPERIMENTAL
CALCULATED
80 -[K]W
300 350 400
Fig.5 Young's modulus vs. temperature (specimens A/a)
Ik
1i'1
i7
Nt-
0 1000 2000
1 E [GPa]
130
120. -
//
~/
1
6
Fig.7 Young's modulus vs. stress (specimens A/a)
E[GPa]
E
80.
60. .6
Ip40. .4
20. ..2
60 0i0 0
"I.
-. ii~!
4-12
E[GP,]
20. "'
10. 10 EXPERIMENTAL
CALCULATED
T[K]
0.
300 350 400
Fig.9 Young's modulus vs. temperature (specimens B i/b)
IV
EXPERIMENTAL
CALCULATED
' ~.75. /
*lfl~.70.
.65 T [K]
300 35o 40o
SbI
"---''"i~
4-13
E~LGP1
18.--- -~ 321 K
Ii; 1- %350 K
363 K
%"
o[G Pal
0 .01 .3 .05
Fig. I I Young's modulus vs. stress (specimens BI/b)
IV 3363 K
LI /./350 I
> ' 3-2 1 K
S.07 .' 293 K
.06.
.. i. .06
([G Pa]
1.0 /
.2.
E[GPa]
200.
180
vII
a [G Pa]
160 lb
293 K
.325K
. I-,
...- --372 K
120-
100-
a [G P,]
.d2 ' i6 .10 "
E [G Pal
I'1 200. -- - - -
CALCULAT ED
I*ii
---- EXPERIMENTAL
150.
tK
' ~T [K'
E[G Pal
100.
CALCULATED
80. EXPERIMENTAL
T[K]
EXPERfMENTAL
.075 CALCULATED
I"
fL '300 350o 400
i'. Fig. IS Poisson's ratio vs. temiperature (specimen C2/a)
B 7
4-17
74
t
Fig.19 Equipment for dynamic tests
4.C
2 e
' 2 3 4 5 6 7 u~v)
Fig.20 Dimensionless damping ratios (') vs. the response amplitude
Ii
.:
" - -
!!
.103
FREE VIBRATION TEST
x SINLISOIDAL FOR CED VIBRATION TEST
X
, I( X X X
O. ' ' (m
1e 16.4 10"2 i 100 ToO P Hg)
,: - ;i 1--- "I
t
q!i)t ---
i.
The dilatometer SFig.22
. ')
Ir -$
!
L',, '
4-19
5..
3. 60
W~ AO
$1 1 20
0 So0 3O 4600
a 10"[K]/
____CALC.
EXP
A 20.
-
Ef10
t=15~
100 200364o
20.
15.
o 120 K
170 K
* 220K
a 240K
10.
5.
IX
0
45'
a o[K]EXP.
CALG.
3C
S400 K
30K
"10.
RECORDER'S REPORT
by
M.F.Card
NASA-Langley Research Center
Hampton, Virginia 23665
USA
Four papers were presented on physico-chemical effects of environment on composite materials. Mr Edge compared
moisture absorption data obtained from aircraft exposure tests with accelerated laboratory tests. Data suggests that in
[Comparisons most cases, moisture pickupmeasuresn
of laboratory in unaccelerated
h tests was
. f accelerated less indicate
testing than one-percent over periods
that immersion of time
in boiling as gives
water long excessive
as five years.
degradatic-n results compared to real life expectations. Humidity and constant temperature exposures appear to give
reasonable results except when maximum moisture content for unaccelerated exposures is exceeded. For ambient
conditions, reductions in strength due to moisture pickup of less than one-percent are small. However, there still is
concern for excessive degradation, particularly when hot, wet environmental exposures occur.
Mr Tennyson presented experimental results on short- and long-term behavior of composites subjected to various
environmental conditions. A space environment test facility was described with provisions for mechanical loading under
vacuum, temperature, solar radiation, and electron or proton exposures. Results were presented for outgassing of com-
posite-epoxy material indicating that water was the principal outgassing component and that in situ testing is needed to
truly characterize material behavior in space. Results of studies of material failure criteria were also presented. The
superiority of a cubic strength criterion was illustrated for biaxial strength predictions as well as prediction of successive
failures in internal pressure tests. Strength and stiffness changes in graphite-epoxy laminates due to vacuum and
temperature charges were presented, as well as analytical predictions, of strength tensor coefficients. Finally, coefficient
of expansion data for graphite-epoxy, Kevlar-epoxy, and boron-epoxy laminates under thermal cycling showed some
changes due to microcracking and outgassing. Predictions of coefficient of expansion changes for angle plies for
ambient versus vacuum conditions were in good agreement with test data.
S Dr Barboni presented experimental and theoretical results for mechanical and thermal behavior of a graphite-epoxy
pultrusion and three types of graphite laminates. The pultrusion exhibited a nonlinear stress-strain curve whose linear
' slope region was insensitive to temperature. Similar behavior was observed in the laminates with the largest temperature
effect observed in measurements of a 90/0/90 ply along a 45-degree angle. Calculated values of Young's modulus and
Poisson's ratio were in reasonable agreement with test. Free and sinusoidal forced vibration tests of the pultrusio'n, in
s and out of vacuum, showed small effezts on ntdulus, but a 15% reduction in ambient damping results. Thermal
r [expansion results for the pultrusion showed gocd agreement between theory and test.
Subsequent general discussions focused on how important moisture distribution was and how it could be more
accurately simulated relative to aircraft exposure tests. Mr Edge reiterated the need for study of tropical (wet, hot)
conditions. The practicality of analytical models for stiffness and strength was discussed. Mr Tennyson indicated that
with known or measured principal values of stiffness or strength, analysis could be used to handle tht variety of laminate
configurations encountered in design.
5-1
F.J. ARENDTS
KO. SIPPEL
D. WEISGERBER
MESSERSCHMITT-BOLKOW-BLOHM GMBH
AIRCRAFT DIVISION
8000 MUNCHEN 80, P.O.Box 801160
W-GERMANY
ABSTRACT
Constant-amplitude and flight-by-flight tests with five different load spectra were done
with unnotched CFRPspecimen. Also the influence of overloads was investigated. The test
results were compared with fatigue life predictions based on "Miner's Rule" applied for
different conditions.
The results show that overloads in all cases cause a significant reduction of the fatigue
life. It turned out tnat fatigue life estimations based on "Miner's Rule" are on the
unsafe side in some cases by more than a factor 10, getting a big scatter among the cases
investigated.
1. INTRODUCTION
The new generation of fighter aircraft will have an increasing amount of CFRP structural
components, which in many cases are subject to fatigue. Whilst comparatively reliable
predictions of the service life of metallic components can now be made with the aid of
damage accumulation theories based on available test results such as flight-by-flight
tests, there is not yet any reliable method of assessing the service life of CFRP
components.
Tests were therefore performed, initially on unnotched CFRP specimens, using the results
of flight-by-flight fatigue tests to investigate to what extent the Palmgren/Miner linear
damage accumulation hypothesis/1,2/is applicable in assessing the service life of CFRP
structures. A further aim was to examine the influence of overloads on the service life,
and to determine the probable scatter of the test results.
2. TESTS
2.1 Test Programme
The exact set up of the layers, and the specimen used for the tests, will be seen from
Fig. 2. The longitudinal axis of the specimens was always in 00 direction, which was
the direction in which the sheets had the maximum strenqth.
The specimens used for tests under part A had an GRP reinforcement bonded in the clamping
area. This requires an additional tempering cycle of 110 minutes at 170 0 C. Since tests
showed that this reinforcement was not necessary, it was omitted from the specimens used
for part B of the programme. An influence on the material properties caused from the
then given omission of the GRP reinforcement bonding cycle does not exist as earlier tests
have shown.
t 2.3 Test Spectrum and Sequences Used in the Flight-by-Flight Fatigue Tests
The flight-by-flight tests were based on the following five typical spectra:
- Fighter aircraft horizontal tail spectrum HLW-1 (No.1) /3/
- Mini-Twist (No.2) /4/
- Fighter aircraft nz spectrum (g-spectrum) (No.3) /5/
- Spectrum for SLAT operation (No.4) /6/
- Spectrum for FLAP operation (No.5) /7/
The typical shapes of the spectra named above are shown in Fig. 3.
For spectra No. 1, No. 2 and No. 3, the influence of overloads was also examined.
It should be noted that, because CFRP composites are so sensitive to compressive stresses,
all loads derived from the spectrum were applied with the reverse signs, i.e. tensile
load cycles became compressive load cycles and vice versa. This was duly taken into
account when placing the plus and minus signs on the ordinate in Fig. 3.
Fig. 8 gives the results for the constant-amplitude load tests, a sketch for regressive
evaluation, and a table showing the relevant constants or boundary conditions.
relating to the service life proved to vary considerably from one load sequence to an-
other, and was particularly high at T = 1 : 16.4 in the case of the n. spectrum tests
(No. 3).
Figs. 9 and 10 also clearly show the influence of overloads. In Fig. 12 the results are
shown more clearly. In the horizontal tail load sequence, the overload G = 1.25 du,
which was interspersed once in every subsequence, has the effect of reducing the service
Vtlife relative to the value PU = 50 % by the factor 3.0. In the TWIST sequence, the in-
fluence ot an overload Zu = 1.22 tu was approximately the factor 1.95. In the nz sequence
4 No. 3, overloads K =1,25 du see also para 2.3.3) caused a reduction in the service life
by the factor of 1.83.
,
ft~ The wide scatter of results which became evident for load sequence No. 3 was what gave
rise to supplementing the programme by part B. Several of the subsequently teste. specimens
were subjected to an ultrasonic test before the beginning of testing and after about 50 %
of the expected service life. One of these specimens, which displayed an obvious ultra-
sonic failure reading before the start of testing, failed after only 91 flights. The
. ultrasonic readings for the other specimens were found not to indicate any obvious faults
S even after half the expected service life.
The points
resultscannot
are shown on a with
probability plot line,
in Fig. 13. means
As is that
evident,
they the values
Sas be joined a straight which do not have shown
a
normal logarithmic distribution. When the distribution was examined for unacceptable
values itwas found that three values failed to meet the assumed confidence inter-
c101n
val of 95 %of the normal logarithmic distribution. The omission of these coordinate
values yielded the distribution resulting from the cross-shaped points, and from this a
scatter value of T = I : 4.9 could be derived. This scatter is of approximately the same
| " order as the scatter occurring in the tests with other load sequences.
"(I
5.4
-2 of the Cvalue from 0.072 to 0.088. Taking the original spectrum as the reference, the
overloads on the Mini-Twist
same sequence should
load tendency by meansbe offound
the when Miner the
assessing
relative rule.influence
Here too,of as can derived from Fig. 16,
the value increases from approximately 0.158 to 0.26 in the case of overload by contrast
to the value without overload. The opposite behaviour - in the case of interspersed over-
loads a relative Miner calculation of service life would yield a result on the unsafe
side - can be expected in the case of the horizontal tail load sequence, because the
C values for cases with and without overloads are 0.336 and 0.408 respectively thus
showing the opposite tendency to that of the Mini-Twist and nz load sequence.
This means that, although interspersed overloads significantly reduce the service life,
the result of the relative Miner calculation does not always tend in the same direction
when the original spectrum in question is taken as a reference.
'rI
5-5
4. CONCLUSION
The investigation revealed that the "Miner Rule" does not yield satisfactory results for
dimensioning carbon fibre reinforced composite materials which will be subjecc to dynamic
stresses in fighter and transport aircraft load sequences. With the exception of the
relative Miner calculation, none of the variants examined was able to improve the accuracy
of prediction to such an extent that variations up to a factor of about 14 were not still
likely. Other investigations conducted by MBB /12/, investigations as described in the
fibres manual /9/ and those conducted by NASA /13/ all showed that results obtained from
the Miner calculation are on the unsafe side.
As expected, a certain improvement in the assessment of service life was obtained with
the relative Miner calculation. However, CFRP seems to be much more sensitive to
differences in load spectra than are metallic materials. Nevertheless, provided that
similar test results are available, this method is at present the best available for
assessing the service life.
However, it must be stated that there is no procedure for predicting the fatigue life
of CFRP components at the moment which is applicable for a designer and proved by tests.
The question if the "Fatigue Residual Strength Degradation Model" by J.N. Yang has the
capability to predict the fatigue life of a component under flight-by-flight loading
deserves further studies. At that time the prediction of fatigue life of CFRP components
must be based on relevant flight-by-flight tests. In Fig. 17 a fatigue life curve based
on the spectrum No. 1 is given. It shows the relative steep slope, which is typical
for compression-loaded CFRP material.
5. REFERENCES
/l/ A. Palmgren Die Lebensdauer von K-igellagern; Z.V.D.J.68(1924) S.339-341
/2/ M.A. Miner Cumulative Damagr Jn Fatigue; J.Appl.Mech., 12(1945)
pp. A 159-164
/3/ P. Keerl, 21ight-by-Flight Belastungsablauf fUr ein Kanipfflugzeug-
D.Weisgerber Hbhenleitwerk, MBB-Bericht Nr. TN-FE 214/3/76
LIST OF SYMBOLS
(N/mm2 ) Applied Stress
FIGURES
ProgrmYn Type of Test Material Type of Specimen load Sequ e/load SpectrZu ofi Te_,st Institute
A Static and dynamic tests with specimen fran one ac plate, for checking Miner's nule
B Tests with CFRP Specimen for evaluating batch influences on the static and dynamic strength
4Static tests CR
CtP Fig.: 2 Tension, 114 DFVLR -
(914C, T300 B) (tmnotched) Coapression 14 Kdln-Po4rz
Flight-by-Flight
fatigue tests CFRP
(914C, T300 B) Fig.: 2
(unnotc~hed) au -600 N/mn', Ioadseq.No. 24 DFVMR -
K?31n-Po.-z
(tIAM
2
FIBRE -ORIENTATION
"-1 ON THE SUFACE
N ~O'DIRECT1ON 1
POS. 1
GRP REINFORCEMENT
.1 POS. 3
LAMINATE
-06
tTYPE NO ' a-g
- - MINITWIST
Mo
0.2 K. HL---i
0.6. I
0.8
I I ' _ ...
': .........
w0 ll ' 0 is1
LOAD CYCLES FOR1000 FLIt,'T HOURS
*f FIG
G LOADqPECTRA USID FOR THE INVESTIGATIONIEVALUATEDACCCROIGTO -HE MODIFIED
LEVEL CROSSINGCOJUT METHCO)
5-8
LOAD P -640N
(k N) -2
4-0 -TIME
20
601
F IG. 6 INVERTED TYPICAL FIGHTER TAILERJN LOAD SEQEEUEE
J1 1 ~AA P~ 1TIME
1.25 1 LC
1.2 LOAD SPECTRU, FOR LOAD aEN3CE NO3
" ----.. /7 LC I"',' N* 0VEF4DADS
0
LOAD CYCLES FOR 200 FLIGHT HOURS
600-
p
[kN] 400,
200
0-0
-200I
-0oo.0
~~ 20C.
Stime
-200
-1100
S' -800
F -1000
- 729,89- 0
0, 4)- 776,38 X = -826,14i Dr___________________
f! 722 S = -69,05
0 - 8.3N ctoad
cycles)
905,2
M ,6 A 0 (1 mal 0,1
916,3
Stress values in N/na' derived by regression
908,5 calculation
o 873,2
_________ 887,7 Fr"
- 40,5
-854,8
2
8092,46 866
8421,4 S
94,
870,7 _...__... ._
- 880
HLW I II WS
3,60
2
13,nr 939 3650 75ml:2 A
so~ i03 0
10 0" o
11 WIHTUJ
OVEROAD
I 941
0.13 0.
5-10
FIGHTER AIRCRAFT
LWAD SEQUENCE NO 3
2 2 2
-600 Nlmm u="6 0 N/mm
G ..:(1,256,u)WITH OVERLOAD .F
.SL LA___P SLAT
00
0
oLi. 50,_
510-
U.10 . _
8t * 1837 3356
* 0.2'0. 0.,t74 *t 9$O I %8
s 0.133 0.388
mlI
lit * 3.3 ? ia.- * I
lIIT * 2.2 9. 87
1111"111l I" I l
1 IIII I
I2
103 NF (FLIGHTS TO FALURE0
101. IIIIII 1 I I II1 1090
1II I II
NF (FLIGHTS TO FAILUREr"
NF ~~ ~ ~ ~(FIHSOFIUE
0 0 0
0
< oo
a) II
s, 0.M
101
,, A;.../
CIO:
i1 , 4
90
12 RESULTS OF FLIGHT-BY-FLIHBT FATIGUE TESTS
I 0FIG.
i !0..
-l N
5-11
FIG 13 RESULTS OF THE FUGHT BY FUGHT FATIGUE TESTS WITH THE nz-
LOAD SEQUENCE NO.3 DERIVED WITH SPECIMEN MANUFACTUFED FRO:
DIFFERENT BATCHES
G
0_
S90-
w
o 50
tL0 g
T I -21.2
p-
co
101
ACCEPTANCE VALUES
x DISTRIBUTION WITHOUT NON ACCEPTANCE VALUES@
i i l l
I II I I -- 1 1' 1 1 I I I I I I I I
SII l I I
a S
~
a-HWI
~SYMOL SPEClRAINO
LOADJ
_- 0.
[N/,''2]
90 0 -6201)
+- SLAT 4 - 620
0 FLAP 5 - 620
T 1.13,7 1) OVERLOAD 1.75 ru
Q MNRjTR 21 OVERLOAD1.22 c
MINER (RFUPTURE)
SIN- CURVEECNSIRING THE 0 NINER IRUPTUR MINERIR'.TUREI
"EMODFlEDLEVBELOSSIG SIN-CURVE. ONSID.-fZN,THE SIN-CURVE WITHOJT CON-
COUNT METHOD ENDURANCE LEVEL SIDERING TIE ENDURANCE
RAINFLOW COUNT METHOD LEVEL (LOG -LOG-PAPER) MODI LEVEL-
-'unsafe safe-- unsafe, =s:f usafe CROSS COUNT METHOD
1.. 1.0
LI,.EA TESTED
FATgUQE
CALCULATED1.
FATIGUE LIFE
..v.
. FIG 14. RESULTS OF THE FATIGUE LIFE CALCULATION
S,3.7 load ~ ~ 30
2,2 A
,1.2
Overloa,l 1,25 u
t.I.T
V,
5-12
1) Overload 1,25~
2) Overload 1,22 c.
(YU
10010 o 0
FIG.~~~~~
LIFE C1R7E
1.1 FAIU
FOWH4~DSEUIE HL -
-70
1111 II -V 1 -~ 11 1 i
6-I1
by
J. J. Gerharz and D. SchUtz
Fraunhofer-Institut fur Betriebsfestigkeit (LBF)
BartningstraBe 47
6100 Darmstadt
Germany (FRG)
SUMMARY
This paper documents a research program on the influence of environment on the fatigue of CFRP unnotched, notched
and bolted specimens. Included are program, details, facility set up, test results, and relevant information found in the
literature. The specimens are simultaneously subjected to load and environmental histories. A flight-by-flight load and
environment sequence typical for the wing-root of a flighter airplane is applied. Test with simplified sirrulation of environ-
ment, allowing high loading frequencies, are run. The admissibility of the simplifications is evaluated by comparing the
results of each simplified test with the results from long time "quasi real time" test with temperatures accompanying th#
loads in each flight. The "quasi real time" flight-by-flight program includes temperature cycles and a humidity cycle. The
specimens are heated and cooled by preconditioned air forced through the test chamber. The resul Is of room' temperature
fatigue tests (base iine data) and of static tests at various environmental conditions are presently available. For a constant
fatigue-stress level residual strength and stiffness data demonstrate the damage growth mode apparent by ultrasonic scan
records.
chapter"-iterature Surve9 many experiments reveal a sensitivity of resin matrix and fibre-matrix interface to both environ-
mental degradation and fatigue loading, with interactions between various mechanisms beeing possible. Assurance is there-
fore needed that any environmental and fatig,,e damage accumulated during a typical aircraft life will not compromise
structural integrit,,. The research program described herein contributes to a base of design data containing the environ-
mental effect on the design properties of the graphite/epoxy composite material.
The application of composites in Europe is presently beeing developed through advanced fighter aircraft. Considering the
aircraft wing a. a primary structure candidate for a graphite/epoxy composite design, the loading and environmental
conditions of the upper surfoce structure were chosen for flight simulation testing. Besides beeing subjected especially to
solar radiation and rain soak, the loading of the upper surface structure is predominantly compressive and graphite/epoxy
composites show a higher sensitivity to cyclic loading in the compression region E 1 ]. A few tests are also planned with
lower surface loading and constant amplitude loading with critical stress ratios.Plain specimens of various laminate lay-up
and jointed specimens with a specific lay-up are us, 1. The wing loading sequence chosen is The FAtigue Loading
STAndard For Fighter Aircraft (FALSTAFF) which is described in detail in [ 2 J. It is based on ctual opercting experi-
ence of fighter aircraft. The average peace ;me usage for this type of aircraft will be 4 flights per week, summing up to
5,33 flight hours/week; thus 1000 flights will be made in 4.8 years, totalling 1333 flight hours C 3 ]. Clearly a real
time replication of environment history in the laboratory to a factored life is not practical, therefore a so-called "Quasi
Real time" te;t program was developed through acceleration of the actual t',mperature/moisture/,oading hiffory. The devel-
opment was jointly conducted with the RPE, Stuctures Department to receive comparable test results from similar RAE and
LBF research programs. The objective of an accelerated test spectrjm should be to develop the same damage growth and
residual strength in the structure as the 'real time loading spactrum but in such a manner that testing time and cost are mini-
mized. By this criterion a major acceleration in real time environmental effects cun i." achieved by shortening the ground
"dwell period, the time the aircrroft spends standing on the ground where composites absorb the most moisture. The average
moisture content and the distribution of moisture over the laminate thickness developing during service life have specific
implications to damage growth and residual strength. This wet condition can be generated and maintained by periodic fast
moisture absorption at higher temperatures and humidities [ 4 1, for example 95 % relative humidity at 70 0 C see tem-
perature profile "C" in Fig. 1 resembling the actual conditions in a tropic climate. Thereby, ground dwell periods may be
shortened drastically by a factor of 400 1 3 ] red'icing total test time from 4.8 years to 8.5 weeks for simulation of 1000
flights. This is still to long and therefore reduction of the flight phases is necessary. Here fatigue loading at extreme term-
"peratur.s, the time for the decay of thermal gradients acro~s the specimen, the heating and cooling rates, the number of
flights with changing extreme temperatures and ,-oisture content have specific implications to damage growth and residual
r strength. Thus, moderate temperature changes can be'neglected and more important, constant tempera,'ure periods need not
to be lon~sr than the time for the load cycles ,occurrio,.a. The FALSTAFF load spectrum applied constists of a repeating se-
quence of 200 flights of vary;ng s.verity. Choosing a malistic frequency oa 5 Hz [ 5 3 for the fatigue loading the most
severe flight will last 56 seconds. With regard to decay of temperature gradients, the period where the extreme temperature
ii constant will last 2 minutes. Keeping the actual temperature change rates of 60 0 C/min, and assuming a mission mix
I I ] of 1/3 of flights with temperatu:e profile "A" in Fig. 1 and 2/3 of flights with temperature profile "B" in Fig. 1
the test time for 1000 accelerated flights including ground dwell becomes %12 days. The derived test spectrum was assumed
to meet the acceleration criterion. T.he test program using this spectrum is called 'Quasi-Realtime" program. This ptogiam
, utilizes both the flight temperature profiles "A" and "B" as well as the ground dwell temperature profile "C" as noted in
Fig. 1. Whereas the flight profiles ocrur corresponding to the mission mix, the ground moisture dweil periods (profile "C")
oc.o'rs once every 100 fliqhts in the accelerated test spectrum.
6-2
As shown in Fig. I additional testing programs are intended to check-out further allowable simplifications reducing
testing time by factors up to 6. The acceptance of these programs is jugded by comparison of their results with those of the
"Quasi-Realtime" tests. The simplifications involve an increase of loading frequency to 15 Hz, omission of a complete tem-
perature profile and of a temperature condition within a profile leading eventually to a shorter moisture recovery period as
pointed out in Fig. 1. For example in the test program called "Fast" the loading frequency is increased to 15 Hz, the tem-
perature profile "B" is omitted and because of that less time is required for moisture recovery.
In any of these testing programs the some loads occur always at the same temperature; if a temperature is omitted the
affected loads will join the room temperature (250 C) periods. The comparison of test results will then show, which of the
environment cycles or environment conditions did not contribute to damage. Fig. 2 gives an example, how t'"e loads of the
FALSTAFF spectrum ore programed to join the temperature cycle. The sequence of temperature and humidity shown, can-
sist- of flight temperature profile "A" followed by the ground dwell temperature profile "C". The sequence of loads repre-
sents the loading for one flight beginning and ending with the groynd loads in tension and with the flight load cycles in
compression between them. Specific rules were obeyed in joining the loading to the temperatuie/humidity cycling:
- Cyclic loading occurs after the appropriate steady state temperature has been achieved; hus temperature and load
does not vary simultaneously. This will ease cumulative damage calculations.
Ground loading is combined with the room temperature periods of the temperature profiles.
.. Constant loads during flight loading, ground loading, and ground dwell periods are close to the mean loads of these
periods.
So far in the test programs mentioned above simultaneous occurence of environment and load has been looked at. But
also testing with flight-by-flight loading separated from flight-by-flight temperature/humidity cycling may lower the costs.
Such testing is part of our investigation accomplished by extensive use of virgin'and prelooded "traveller specimens". The
outcome of these experiments will give some answers to the question whether or not the cyclic loading correlates with envi-
ronment cycling in their effect on damage development.
In summary the experimentoa %Yorkof our research program can be grouped as follows:
- flight-by-flight loading at room temperature, baseline testing,
- accelerated flight-by-flight temperature/humidity cycling simultaneous with flight-by-flight loading ("Quasi Realtime"
testing, baseline to simplified programs),
- simultaneous temperature/humidity and load cycling omitting flight temperature profiles and temperature conditions
(simplified programs),
- flight-by-flight loading seperated from flight-by-flight temperature/humidii, tycling (applies to "Quasi-Realtime"
and simplified programs) and
- flight-by-flight temperature/humidity cycling only.
TEST SET-UP
All the fatigue tests and the static and residual strength tests at high and low temperatures are conducted in eight load
frames. In each load frame two specimens are loaded in parallel by one 20 kN-capacity hydraulic cylinder. Four locd
frames are blocked together. The test sections of the eight specimens of each block are sitting in a channel through which
the preconditioned air is blown with about 10 m/s. Thus 16 specimens can simultaneously be subjected to temperature/
S', moisture/loading. A general view of the test set-up is shown in Fig. 3. The two blocks of load frames are central in the
picture. A waikway between the blocks gives ready access to specimens on both sides of the blocks. On the left side some
"ofthe specimen-temperature control and recording equipment can be seen. Behind this the tower of cold and hot chambers
(designed by Weiss Technik GmbH) shows in which the air is preconditioned to selected temperature and humidity levels.
"Thepreconditioned air is circulated through the insulated ducts over the test sections of the specimens.
half A close-up view of four blocked load frames is shown in Fig. 4. From the channel guiding the preconditioned air one-
half is removed to exhibit the specimens, clamped between antibuckling guides. In the b'ackground part of the Schenck
analog-load control equipment shows, seen completely on Fig. 5. On top of the central rock is the control equipment for
the rate of temperature change. The visual display and typewriter terminals in front ar* used to put in the environment
programs and test parameters, to start the test and to report specimen failure respectively. The twelve channels of the x-t
recorders are used to continuously document the measured deformations and the corresponding load signals from eight load
i' 1 cells.
Loading Equipment
intA 32 K words memory process computer was programed to read the flight-by-flight loading sequence as command signals
"into the system shown schematichly on Fig. 6 (Load and Environment Programmer). The loads are controlled by Schenck
analog load controller. Calibrated strain gages on the piston (load cell) in the hydraulic cylinder provide a feed back
S..signal to the analog load controller to produce an error signal for correction.
One of the eight load frames is also shown schematically in Fig. 6. The load is carried over from the hydraulic cylinder
*--t Z
6-3
to a stiff beam linked on both ends to the lower fixtures of the specimen, and to the hydraulic brakes. The lower specimen
fixtures are connected to a calibrated linear transducer by which the specimen deformations are measured. When the speci-
men deformations exceeds a present value, for example at specimen fai!ure, a signal from the system controller activates
the hydraulic brake linking that end of the beam where the specimen failed rigidly to the load frame. Simultaneously the
specimen failure signal is given to the process .:omputer causing the teletyp to report the failure data. Loading autnmatically
continues to failure of the unbroken spe -imen; when this fails, a signal shuts off tho pilot valve, depressurizing the hydrau-
lic load system.
The sume process computer providing the load command signal was programed to supply the signals to the temperature/
humidity control system shown schematically in Fig. 7. The signals change temperature and humidity by a preset rate at a
defined !oad cycle within the load sequence of the flights, see Program Description. The timing of the temperature cycling
ardthe loads is realized by a special command.
Thermocouples in the specimen test chamber feed a Temperature Controller from where the process computer expects a
signal to continue loading, w!,:.r the temperature levels of the environment rogrom are reached and a WAIT command was
used to interruct cycling looding during temperature clringes. Cooled (- 80 C), heated (+ 130 0 C) and moisturized air
(95 % r. h. at 70 0 C) are generated in two chambers. Temperature cycling of the chatber enclosed specimens is realized
by changing the voltage of the heater and by opening and closing flaps in the ducts connected with the specimen chambers
through which the preconditioned air is forced by blowers. The voltage c'iaige for the heater and the blower speeds are
controlled to meet preset temperature change rates.
TEST RESULTS
Of the research program presented above we hove so far carried out some baseline testing with plane specimens frorna a
graphite/epoxy laminate with a [ * 45 1 S 8 lay-up These experiments include static tests at various environmental
conditions, FALSTAFF flight-by-flight, and residual stiangt.%testing at room temperature. The * 45 0 -laminate was
chosen to start with because the influence of temperature :nd humidity will show up very clearly as the affected matrix and
interface is highly stressed throughout the total specimen cross section. Those of the environmental testing programs men-
tioned before which will not influence damage growth and residual strength of this laminate, may be excluded from the
investigation of 00 / _ 45 0 laminates where the portion of highly stressed mutrix and interface is less.
The program is behind schedule due to difficulties with the environment testing equipment which is still worked on.
All specimens are made from the same prepreg system, consisting of 60 percent vol. of HT fiber T 300 from Toray and
epoxy resin 914 C from Gba. This material was selected, because preceding investigations have already provided a large
amount of experimental fatigue data on plain, notched and jointed specimens fabricated from this material [ 37, 38 3.
Furthermore the RAE , Structures Department, the DFVLR*, Braunschweig are using this prepreg system in their inves-
4 tigatioriinta the effect of environment.
From the T 300 / 914 C Tape prepreg material the LBF manufactured, in an autoclave (curing temperature 170 0 C),
unifr.m plates of the 18 layer angle ply composite ( [ + 4 8 S) from which specimens were cut immediotly after the
curing process. As the cut was wet, the specimen were dried in an oven at 50 C for half an hour before they were
J numbered and weighed and then sealed in a plastic bag and stored in a freezer.
The specimen I 3ve a constant cross section and unreinforced gripped areas, Fig. 8. An anti-buckling guide prevents
compressive buckling by supporting a 4 mm wide area, central to the longitudinal axis of the specimen, as shown in Fig. 8.
This designi does not influence the development of damage which starts at the free edges of the specimen and keeps the
"specimensu-face accessible to the environment. The clamped-on anti-buckling guides allow a specimen deformation of
4 mm during compressive loading. This type of specimen and anti-buckling guide were applied in static and fatigue tEsting.
The linear transducers fixed to the lower-grips of the testing machine measure the longitudinal extension of the spec -
* mens as mentioned before. These devices were calibrated by simultaneous measurements with strain gauged extensometers
installed on the specimen. Deformation was continuously recorded during monotonic and cycling loading; the loading rote.
were 50 N/s and 15 Hz, respectively. During static tests at + 110 0 C and - 50 0 C the temperature was controlled by
a traveller specimen with embedded thermocouples showing that the specimen pick up the temperature within a few seconds.
All specimens for residual strength testing were C-scanned before and aftei fatigue loading. The specimen were gripped
between smooth jaws, damped on by controlled torquing of the screws, to provide a gradual load transfer to the specimen.
Experience in composite fatigue data analyses has shown that static test results were a needed part of this information.
Static test data are assembled in Table I. It summarizes test results of wet and dry specimen- loaded at R.T., + 110 to
115O0C and - 50 to - 550C.
As fatigue loading with FALSTAFF flight-by--flight loaded sequence for wing upper surfaces causes damage growth and
total failures to occur at compression, static testing was done in compression, too. Plastic shortening of the specimens to
the limit of the test fixtures terminated the compression test. The average corresponding strain was three percent. Because
no fra:ture but lateral bending occured at large strains, the stress measured at half the limit strain was chosen to represent
compressive strength and stiffness. The rationale for this choice is the occurence of fatigue fracture when increasing defor-
matior, approaches 1 .5 o% strain, as demonstrated in the following chapter, and in 1- 45 0 graphite/epoxy laminates, the
same modes of failure occur in static compression an in compression-compression fatigue loading as elaborate scanning
electron microscope observations by G. C. Grimes [ 39 1 have shown.
Fig. 9 shows the stress-strain behaviour at various environmental conditions. The influence of humidity is, for reasons
of clarity, separately illustrated in Fig. 10. Generally all stress-strain curves have a non-linear pattern with a relatively
small, not well defined linear portion at the origin. Therefore the compressive modulus E c is determined from the secant
line drawn between the origin and the load deformation point at 0.1 % strain C 40 3.
A striking feature is the considerable influence of temperature and moisture that shows up. Wheras at + 115 0 C and
in the wet condition (at + 115 0 C and room temperature) strength and stiffness has drastically reduced, at - 50 0 C the
opposite is observed; stiffness and strength have become larger and moisture content seems to increase stiffness even further.
Fig. 10 shows that the detrimentao influence of moisture content is largest at + 115 0 C and much smaller but still
significant at room temperature.
From the static test results and results of fatigue test with the FALSTAFF load spectrum C 1 3 a stress level was
estimated which would result in fatigue lifes ranging from 104 to 2 x 104 flights. At the stress level, S = - 137.5
N/mm2 , specimens failed in the average at L = 1.3 x 10 4 flights. Therefore this stress level was chosen to be used for
further testing. Accordingly the specimens for the residual strength tests were preloaded on this stress level, too.
Lifes to failure and lifes to the end of test without failure are summarized in Table 2. Failure as well as non-failure
data were utilized in the statistical analysis. For the high censored samples point estimates of the Weibull parameters
04. and 1 were determined by the maximum likelihood estimation. The mean life E (L) and the standard deviations
listed also in Table 2 were calculated through analysis of log. life values. The listet test results are also plotted in an
SN-diagramm, Fig. 11. The SN-cu, is fitted to the mean values of log. fatigue life at the two stress levels. This is the
SN-curve for 50 percent survival if !og.-normal distribution is assumed to fit the data.
All specimens failed within the gauge length, as shown in Fig. 12. As mentioned in the description of the loading
facility, the fatigue loading of specimen is stopped wben it's deformation overshoots a preset value. For two specimens this
hapnend before fracture occured. Both specimens showed local debonding and delamination at their edges and it is assumed
that continued fatigue loading would lead to tctal failure within a negligible fraction of total life, as experienced many
times before. As already mentioned the deformation produced during fatigue loading of specimens was registered continuously
by linear transducers. From the recorded specimen deformation the values of the ratio E / E00 were determined, where
"isthe strain range at any flight and E go is the strain range at the first flights. y a protcdure .rughly outlined
in Fig. 13 the corresponding ratio of compressive secant moduli E / E s at the largest compressive loads were
calculated. In Fig. 14 this stiffness ratio is plotted against the percentage oJ number of flights relative to number of flights
at failure. All data lie within a narrow scatterband demonstrating that stiffness decreases sharply towards the end of life.
Just before fracture the remaining stiffness is '70percent of the initial stiffness at the beginning of fatigue loading. At the
stiffness ratio E s / E s, o = .66 a horizontal line is drawn to indicate the compressive strength criterion. It was
I
,"iinitial
assumed that the specimens nctually failed when stiffness had dropped to 2/3 of the initial stiffness. With the maximum
compressive strain c, o = - 1 .0 % the corresponding strain at fracture is Ec = 1.5 %o.
Investigations of stiffness or deformation of composites behavior under fatigue loading have shown that the appearing
stiffness decreasgapparently corresponcdtodamage development. The residual strength was investigated after preloading to
6 400 and 9 600 flights. 6 400 flights where chosen because from this number of flights on the decrease of stiffness had
become discernible. The 6 400 and 9 600 flights correspond to 50 and 75 percent of the mean life to fracture, respectively.
The "compressive strength criterion"was also used to determine residual strength values from stress-strain curves of the
preloaded specimens. Table 3 summarizes the results of residual strength test. Besides the data of specimens preloaded to
9 400 and 6 400 flights on the selected stress level S - 137.5 N/mm 21 other residual strength data for various lifes and
stress levels were listed. For the preload stress level S = - 137.5 N/mmi the residual strength values were plotted in Fig.
15 versus number of flights of the preloading. The corresponding lifes to fracture and the compressive static strength show
also on this diagram. The static strength values were plotted on a vertical line at N = 10 0 and the lifes to fracture on a
6-5
horizontal line at S c S
Curves of constant probability of survival are drawn combining the mean values, the 10 pe cent survival, and the 90
percent survival data calcu!oted at constant lifes for the static and residual strength data and at constant stress for the life
to fracture data. The residual strength for 10 and 90 percent ojrvival and for the mean ( E (S) ) were computed using the
equations in Fig. 15. The statistical analysis applies the two parameter Weibull distribution and considers the facts that
specimens fail in fatigue when their residual strength is reduced to the fatigue stress level and that the residual strength
data stem from surviving specimens. The same analysis method has been used by J. Awerbuch in C 41 1.
The residual streng:t curves show that the strength of * 45 0 -l aminates decrease long before fatigue failure occures.
In contrast to this behaviour the strength of unidirectional laminates loaded in the direction of fibers decreases rapidly
immediatly before fatigue failure [ 41 ]. The shape of the curves resembles that of the stiffness decrease curves. C-scan
records of the preloaded specimens shown in Fig. 16 reveal discemable delaminations. Specimens loaded monotonically to
tensile fracture actually broke at locations where ultrasonic C-scanning exhibited this damage, see Fig. 17. In Fig. 18 and
19 the stress-strain curves and C-scan records of prelooded specimens show how the damage, identified by ultrasonic C-
scanning, correlates with the degradation of strength and stiffness at static tensile and compressive loading, respectively.
The influence of prelooding is apparently the same in tension and compression. However, the shape of the stress-struir.
curve of the specimen without any delamination,as indicated by the corresponding C-scan record, suggests that another
mode of failure preceds delaminat'on. This has been observed by J. Grimes C 39 ] with a scanning el ,ctron microscope
in -.1 45 o graphite/epoxy laminates which hod also been fatigue loaded in compression - compression. ie concluded that
the modes of failures which occured were debonding of fibers at ply interfaces followed by ply delamination and occasional
fiber microbuckling. Thus deboasding not detectable by the ultrasonic technic seems to be responsible for degradation when
loading in compression has not yet caused delamination.
MOISTURE CONDITIONING
Moisture content as well as moisture distribution through the thickness of the specimen affects the hygrothermal state
of stress in the composite laminate. Therefore it is desirable to not only keep the moisture content on a realistic level but
also to have realistic moisture distribution during flight-by-flight loading. To be on the safesideit was assumed that the
wing structure will level
home-base of the aircrafts is located in very moist area; then the moisture content in the compositehave
at one percent [ 8 , 9 , 18 1. Except in room temperature base-line testing all specimens will been preconditisoned
to at least 1 % moisture content. The distribution of moisture corresponding to realtime exposure will be predicted using the
computer program W8GAIN developed by G. S. Springer C 4 1. Therefore about 50 specimens have been placed in an
environmental chambnr to pick up moisture. Five groips of 10 specimens cut from 5 different plates were put in the chamber
at different times. After different periods of steady state conditioning at + 65 0 C and - 98 % r. h. the groups were taten
out and each specimen of the group was weighed. There was only a negligible small scatter in the data of weight increase
within the groups of specimens. The average weight increase of the ten specimens from one platc were plotted in Fig. 20
against the square root of exposure time. By fitting the weight increase function derived from Fick's law to the measured
data, the diffusivity and maximum moisture content presented in Fig. 20 were established.
The moisture recovery times needed during flight-by-flight testing to keep the taoisture content on the required level
- Iwill be estimated by the program WBGAIN using these and additional data from the analysis of experimental moisture
absorption presented in [ 42 1. Whereas the moisture content of the specimens during environmental fatigue testing will
be controlled by measurements on traveller coupons accom-anying the test specimens, a computational method is applied
to monitor the distribution of moisture.
'4
:In
LITERATURE SURVEY
the literature on mechanical behaviour of graphite/epoxy composites only few information was found which is
related to testing with flight-by-flight temperature/moisture c6c'nges applied simultaneously with flight-by-flight Ioading
C 6 1, [7 , [ 8 1, [ 9 ], [ 10 ], [ 11 ]. Some fatigue and residual strength data from flight-by-flight load/
I, - environment fatigue testing are reported ;n [ 7 ], [ 8 J and E 11 1. Besides providing design data, the investigations
,ererred to in [11 1 are part of the AFML's "Advanced Composites Serviceability Program" using an accelerated test
spectrum related to the temperature/misture/loading history of the B-1 Vertical Stabilizer. No such testing is known which
simulates the temperature/humidity/loading history at a wing upper surface structure. Table 4 summarizes briefly some of
the research work on the influence of temperature and humidity (combined with cyclic loading)on residual and fatigue
strength of graphite/epoxy laminates.
, The test results indicate only minor influences of environment when stressing is predominantly tension but larger
influence when stressing is predominantly ccmpression, shear or bearing.
- Howevei; the vast amount of literature, reporting on effects of temperatu, md humidity on static and fatigue strength,
show that at least temperature cycling, specifically in the low temperature range, E 12 3, E 13 1, C 14 3, [ 15 3
as well as high temperatures during loading and moisture contents larger than 1,0 % degradate mechanical properties
16 J, [ 17 ], 418'J, [19).
- A neneral conclusion is, that degradation is largest for lay-ups and loadings with inherent high stressing of the matrix
and/or the fibre matrix interface [ 12 3, [ 20 1, E 21 3.
Physical approaches to understanding of these phenomenons have shown that they are related to:
6-6
- residual stresses due to thermal anisotropy ( 13 J, [15 3, [ 22 ], [ 23 ], [ 24 3, [ 25 3,
- residual stresses due to swelling with moisture pick-up [ 23 3, [ 26 3, 1 27 3,
- decrease in glas-transition temperature with increasing moisture content [ 92 3, [ 343 1, [ 528 1, [ 670 1, and
- matrix crazing [18 1, [ 31 ].
CLOSING
With the wing upper surface load/environment history accelerated for experimental s0-nulation, still missing design
data will be established. Initial test results of this research program show that:
Sstress-strain behaviour of .45 8 s -graphite/epoxy laminate at compression is strongly influenced by temperature
and moisture content,
- heating to + 110 0 C and / or moisture content of about 1 .5 % weaken the composite laminate whereas at the .reezing
temperature of - 55 0 C the laminate had stiffened.
The results of room temperature fatigue and residual strength testing with the FALSTAFF flight-by-flight load sequence
indicate that decrease of stiffness and strength correspond to damage growth as made apparent by ultrasonic scanning.
The most essential facts pertinent to the mechanical behaviour of composite material under temperature/humidity/
loading conditions were extracted from the literature; they are:
- Generally, degradation is largest for lay-ups and loadings with inherent high stressing of the matrix and/or fibre-
matrix interface.
- With cyclic loading predominantly in compression the influence of environment is larger than with cyclic loading
predominantly in tension.
Temperature cycling, specifically in the low temperature range, high temperatures during loading and moisture content
above 1.0 % degradate mechanical properties of graphite/epoxy composites significantly.
ACKNOWLEDGMENTS
The research program on the "Fatigue strength of graphite/epoxy composite material" was sponsored by the German
Ministry of Defence Research Branch R0 Fo IV. The "Fatigue strength of graphite/epoxy composite material" working
party was entrusted with planning and monitoring the program; the working party was brought into being by the German
Defence Ministry and consists of members of Doinier, MBB, VFW-Fokker, DFVLR-Stuttgart, BWB-ML and LBF.
REFERENCES
E11 SchUtz, D. and Gerharz, J.J.: Fatigue Strength of a Fibre-Reinforced MaterialComposites, October 1977.
t
C2 3 Laboratorium fur Betriebsfestigkeit, Darmstadt, Ed., March 197& Description of a Fighter Aircraft Loading
STAndard For Fatigue Evaluation - FALSTAFF - Flugzeugwerke Emmen (F+W), Switzerland, Laboratorium fur
Betriebsfestigkeit (LBF), Darristadt, Gennany, Nationoal Aerospace Laboratory (NLR), Amsterdam, Netherlands,
Industrieanlagen-Betriebsgesellschaft mbH (IABG) Ottobrunn, Germany.
( 3 ] Webb, J.N.: Environmental Fatigue Testing of Composite Structural Elements, RAE, Structures Department,
30. Nov. 1978.
C4 3 Springer, George, S.: MAoisture Content of Composites Under Transient Conditions, J. Composite Materials,
Vol. II, Jan. 1977, p. 107.
[ 5 ] Cardrick, A.W. and Smith, M.A.: An Approach to the Development of Meaningful Design Rules for Fatigue-
Loaded CFRP-Components, Composites, May 1974.
..4[ , 6 No. 77-464,G.;/March
Waggoner, f H.: D
Erbacher,
1977. Tolerance Program for the B-1 Composite Stabilizer, AIAA Paper,
o':noge
[7 1 Haskings, J.F.; Wilkins, D.J. and Stein, B.A.: Flight Simulation Testing Equipment for Composite Material
Systems. In: Envirenmental Effects on Advanced Composite Materials, ASTM STP 602, 1976, pp. 23-36'&
.1[8 1 Wilkins. D.J.; Wolff, R.V.; Shinozuka, M. and Cox, E.F.: Realism in Fatigue Testing: The Effect of Flight-
by-Flir~ht Thermal and Random Load Histories on Composite Bonded Joints. In: ASTM STP 569, 1975, pp. 307-
322.
I 9 3 l.undemo, Chr. Y. and Thor, S.E.. Influonce of Environmental Cycling on the Mechanical Properties of
Composite Materials. J. Composite Materials, Vol. 11, July 1977.
#-i
6-7
1 10 1 Konishi, D.Y. and Johnson, W.R.: Fatigue Effects on Delaminations and Strength Degradation in Graphite/
Epoxy Laminates. NA-78-380, 1978.
C 11 1 Altman, J.M.. Advanced Composites Serviceability Program Status Review. North American Aircraft Div.
Rockwell Internation, Los Angeles, 1979.
C 12 ] Hofer, K.E.; Larsen, D.; Humpreys, V.E.: Development of Engineering Data on the Mechanical and Physical
Properties of Advanced Composite Materials. III Research Institute, AFML-TR-74-266, Febr. 1975.
13 J Mazzio, V.F. and Mehan, R.L.: Effects of Thermal Cycling on the Properties of Graphite-Epoxy Composites.
In: Composite Materials: Testing and Design (Fourth Conference), ASTM STP 617, 1977, pp. 466-480.
C 14 ] Camahort, J.L.; Rennhack, E.H. and Coons, W.C.: Effects of Thermal Cycling Environment on Graphite/
Epoxy Composites. In: Environmental Effects on Advanced Composite Materials, ASTM STP 602, American
Society for Testing and Materials, 1976, pp. 37-49.
[ 15 3 Daniel, I.M. et.al.: Lamination Residual Stresses in Fiber Composites. III Research Institute, Chicago,
Illinois, March 1975, N 75-30264.
16 ] Chi-Hung Shen and George S. Springer: Effect of Moisture and Temperature on the Strength of Composite
Materials. J. Composite Materials, Vol. 11, Jan. 1977, p. 1.
E 17 ] Chi-Hung Shen and Springer, G.S.: Environmental Effects on the Elastic Moduli of Composite Materials.
J. Composite Materialf, Vol. 11 (July 1977), p. 250,
[18 1 Hedrick, J.G. and Whiteside, J.B.: Effects of Environment on Advanced Composite Structures. AIAA Paper
77-463, AIAA Conference on Aircraft Composites: The Emerging Methodology for Structural Assurance,
Son Diego, Calf,/ March 24-25, 1977.
C 19 2 Weinberger, R.A.; Somoroff, A.R. and Riley, B.L.: US Navy Certification of Composite Wings for the F-18
and Advanced Harrier Aircraft. AIAA Paper, No. 77-466, March 1977.
C 20 ] Kalble, D.H.; Dynes, P.J.; Crtne, L.W. and Maus, L.: Kinetics of Environmental Degradation in Graphite-
Epoxy Composiles. In: ASTM STP 580, 1975, pp. 247-262.
[ 21 Judd, Nigel C.W.: The Effect of Wateron Carbon Fibre Composites. In: Reinforced Plastics-Milestone 30,
Proceedings of the 13 th Anniversary Techn. and Mana,ament Conf., Section 18 A, 1975.
f 22 2 Trabocco, R.E. and Stander, M.: Effect of Natural Weathering on the Mechanical Properties of Graphite/
Epoxy Con.posite Materials. In: Environment Effects on Advanced Composite Materials, ASTM STP 602,
American Societ,, for Testing and Materials, 1976, pp. 67-84.
C 23 2 Hahn, HT.: Residual Stresses in Polymer Matrix Composite Laminates. J. Composite Materials, Vol. 10
(Oct. 1976), p. ?66-278.
r 24 2 Koelble, D.H. and Dynes, P.J.: Nondestructive Test for Shear Strength Degradation of a Graphite-Epoxy
190
Composite. In: Composite Materials: Testing and Design (Fourth Conference), ASTM STP 617, 1977, pp. .
C 25 ] Daniel, I M. and Liber, T.: Lamination Residual Strcins and Stresses in Hybrid Laminates. In: Composite
Materials: Testing and Design (Fourth Conference) ASTM STP 617, 1977, pp. 331-343.
C 26 1 Halpin, John C.: An Assessment of Life Assurance Methodology for Advanced Composite Structures
Unpublished Paper (AFFDL, Wright Patterson AFB, Ohio).
.27 Pipes, R.B.; Vinson, J.R. and Chou, T.W.: Ohn the Hydrothermal Response of Laminated Composite Systems
"L2 2 J. Composite Materials, Vol. 10. April 1976, p. 129-148.
E 28 2 Browning, C.E. and Hartness, J.T.: Effects of Moisture on Properties of High-Performance Structural Resins
"andComposites. In: ASTM STP 546, 1974, pp. 284-302.
[29 ] August, A.; Hodcock, R., and Dastin, S.: Composite Materials Design from a Materials and Design Perspective
I C Grumman Aerospace Corporation, Bethpage, New York Presented at the AGARD Fall Meeting, 1975, Ankara,
r 30 1 Roylancee, D. and Roylance, M.: Influence of Outdoor Weathering on Dynamic Mechanical Properties of
Glas/Epoxy Laminate. In: Environmental Effects on Advanced Composite Materials, ASTM STP 602, 1976,
pp. 85-94.
C 31 2 Browning, C.E.; Husman, G.E. and Whitney, J.M.: Muisture Effects in Epoxy Matric Composites.
In: Composite Materials: Testing and Design (Fourth Conference), ASTM STP 617, 1977, pp. 481-496.
S32 ] Sumsion, H.T. and Williams, D.P.: Effects of Environment on the Fatigue of Graphite-Epoxy Composites.
In: ASTM STP 569, 1975, pp. 251-261.
C 33 2 Hofer, K.E. Jr.; Bnnnett, L.C. and Stander, M.: Effect of Moisture and Fatigue on the Residual Mechanical
Properties of S-Glas/Graphite/Epoxy Hybrid Composite. lIT Research Institute, Chicago Illin., Naval Air
Systems Command, Washington D.C., 1976.
34 1 Bevan, L.G. and Sturgeon, J.B.: Fatigue Limits in CFRP. International Conference on Carbon Fibres, their
Place in Modem Technology, London, February 1974, The Plastics Institute, Paper No. 32.
E 35 2 Bevan, L.G.: Axial and Short Beam Shear Fatigue Properties of CFRP Laminates. Composity j, Vol. 8,
ji Number 4, Oct. 1977, pp. 227-232.
F 6-8
E
36 ) Sendeckyi, G.P.; Stalnaker, H.D. and Kleismit, R.A.: Effect of Temperature on Fatigue Response of Surface
Notched / ( O/+ 451S3 Graphite-Epoxy Laminate. Paper presented at the ASTM-Symposium on Fatigue of
Filamentary Coposite Materials, Deniver.
E 37 ) Gerharz, J.J.; Rott, D. and SchUtz, D.:Schwingfest.1h.Jitsuntersuchungen an ungekerbten und gekerbten
Faserverbundwerkstoffploben aus multidirektionalem Laminat. Forschungsbericht aus der Wehrtechnik
BMVg-FBWT 79-25, 1979.
E 38 3 Gerharz, J.J.; Rott, D. and SchUtz, D.: Schwingfestigkeitsuntersuchungen an Fugungen in Faserbauweise.
Forschungsbericht aus der Wehrtechnik, BMVg-rBWT 79-23, 1979.
" 39 ) Grimes, G.C., and Adams, D.F.: Investigation of Compression Fatigue Properties of Advance'd Composites
Northrop Technical Report, NOR 79-17, Oct. 1979.
r 40 2 ASTM Committee D-30- Standard Test Method for Compressive Properties of Oriented Fiber Composites,
ASTM Designation: D 3410-75.
C 41 2 Awerbuch, Jonathan and Hahn, H.T.: Fatigue and Proof-Testing of Unidirectional Graphite/Epoxy Composite
In: Fatigue of Filamentary Composite Materials ASTM STP 636, K... Reifsnider and K.N. Lauraitis, Eds.,
1977, pp. 248-266.
[42 3 Loos, Alfred, C. and Springer, George S.: Moisture Absorption of Graphite-Epoxy Composites Immersed in
Liquids and Humid Air. J. Composite Material, Vol. 13, (April 1979), p. 131.
~i I rk
I, ..
t--
c
0
0 C!
2' E NnC)
00 0 0
* 's N
V 20
" "
N
" ,c C.) 4
00~
C4.
'0 I"--
E-o N C" N
- N"
+3~ 0
1)N0 N
I:!I 0 a U E
oo o
n4-,
M E~~~ ~ ~~ .. .
0
EZ W
C4 N )fC40 't1 n C N 0 * r0
2 C NCS4
M C* N4C4NC4 CNN -N -- -4
0-
Itn.
CC
CO.C)'
V 0 4cpco &no o o m U - a
01 10O*0C'.
Coco.' 01 *00 a o i
Ir
6-10
I
A.o
bE .. )-
.
expected value of
4i
-.77 i7~~
6-1
o 44
U- U)
. - 0 0
U-o .2
e 0 0
"2 N
+1'0U 2
E1 0O E~'
"I CC 6c C
U) !Q 0 D 0,,
00
C1 .2w LL Vr 0
- U I %0 lN
J- 0j C.
0 m DL m 10 N0 n o oo o o
41 _22 _-- - -_ -_ --
c 0-o
.1 (A p)-p a otao
0 E
ID
0 "n o n oc4 og o No o oL
1 IL 00
V)
F41 C-4 N- 10 F4
6-12
0 0
-u-
0 0 C14a
E 282 E
Lon-
-c-
C n .9 0)
0.40- 0
*u
0 0 0
- 000u000_0
Ea 0.2 0 0 0c 0c
o, 0,
ii0 U 1.4-
0 u -0 )0).
.- 4) 0j .2 .
Du E >, 0>
00 200 22 u0
ku-u
2
2.-u
0a C
E 2 0 .
0g -- 0 -0
a 0 E 00.
c2 0
a0 0 >
0 00 c.-
IE ; 'I_ to
0o e
2 00
E2 02 'Do2
u0 0)
J . 0 1- o-0 0
N~ m- +..' .2 1 .
o o)
0 1
CC
>; -. j 4
co)1
:) 4) - -0
6-13
0 I
a
E E
0)a S SE0
E- 3 )n ) 3 m D0
- 4 0 '0 o
-. C -0 -C -
o0
C 0 a0
ti 0
C0_Xn 0-
4 - 0 0
- - a
C
0 0)
o. C4 C a .2)
:0 10 - )
_jO 4) Q
0~~ 0) C ~ * ~ 0~~-
0- 4V )4)4 M
.2~ 2 .2 L20
.00. 0
C0 C
4i 0 -+ Z 0)
0 0 0 000 04
4) 0-
CL C
C- 0) 0)0 0) >
2 .2x=E x *aE i
-'
2O ,-04)
cn o.
C 00 0 0 0 E
Sn . 0 c 0
0!
0 '0 u0 z
.~. -i 0
C'
Ln .. 0 4.1-
6-14
C MR Time Less
+700C Than in (D)1
S(\25*
"Moisture" --- B
Relat. Humidity
-20 C
-Sequence
of
Temperature
and Humidity
--,- During Flight-Operation P *- on the Ground-
4 1i Sequence of
Load
Time
-'A
6-15
lift
'--M
I
:~ Deformation
- Broke
W2 I i
e-Pesr
11-
C i
6-17
Llowd FramSieIt'
Coninu eoa
ArISupredAe
Iia av
Pressur
'Return
_16
f [~.-100
we
+115
. dry
. RT {dry -200
dry
., ' '-so*{ wet
-300
Curves Represent Average of at Least 2 Spacimens
140
Fg 12 Specimen F d in F
IL
o 4
FeSod
V ID
Pe d10 E <
0N . - .
L; E
z r
I 0; 06
K
Fig. 13
- Determination of Stiffness Ratio from Measured Strain Data
'iE OE- Ea-p toW"]
0 10 20 30 40 50 60 70 80 90 100
* Strength Data
E 160
"" ~10
~It
6-22
,
218
139
117
~ I 160
173
- WOMU 212
I 187
114
196
~ 116
II 11IIIE208
Loaded to 91400 Flights
Fig. 16 - C-scan Records of Specimens
6-23
Load
F.R 9400 FALSTAFF-Flights at Sm-137.5 N/mm 2
7500
Virgin -Specimen
Pre!oaded -
Specimen'
,! 5000-
If 2500-
0 10 20 30
- Strain C. e/s.
Strain E., %. 30 20 10 0
I I 0
-2500
- 5000
I ereloaded
Specimens*
Virgin Sipecimen
-7500
1.8-
:A1.6- :
S~0
j 0 PLateNo. 32
": 03 PtLteNo. 28
0.6,A PlateNo. 31
A0- PLate No. 29
0.40 PlateNo. 30
jj 0/ 5 10 15 20 25 30 35
12
Quare Root of Time, ft, h
(j ~i 8S_
7.
FATIGUE TEST RESULTS OF CARBON FIBRE REINFORCED PLASTIC F 28
AIRCRAFT COMPONENT AND ITS STRUCTURAL DETAILS.
SUMMARY.
As a part of the development program on the structural application of carbon fibre reinforced plastic
composites in aircraft structures, a speedbrake was designed and built in CFRP and aramide fibre composites
for the Fokker F 28 Fellowship. The program was divided into three phases:
- The first phase consisted of a study of different configurations, from which one was selected for a detailed
design in the form of a schematic model.
- In the second phase, tesLs were carried out on structural details and two :chematic speedbrake models were
fabricated for full scale static and dynamic load tests.
- The third phase contained the production of one speedbrake for flight testing on an operational aircraft.
The tests resulted in a design of the speedbrake in which no metal parts were used and in which all joints
were bonded with 120 C and room temperature curing adhesives, without the use of additional fasteners.
The first model speedbrake was successfully tested to ultimate load without failure or any plastic defor-
mation. The second model speedbrake was tested at a maximum fatigue load equal to 615% of the ultimate load.
The full-size CFRP-speedbrake showed a weight saving of 25% when compared with the Al-alloy design. Flight
testing under operational conditions with the NLM started in November 1978.
The application of advanced composite materials is considered as an important contribution to the weight
saving of future aircraft structures. A large number oi investigations has been done to all kinds of para-
meters such as: material properties, fiacture mechanics and durability. This has resulted in a considerable
know-how of material behaviour. The developments in the last couple of years make it possible to design and
build components with controlled properties and durability.
A number of projects were started at the Technological Centre of Fokker to design, test and manufacture a
number of aircraft components with increasing difficulty. The first project concerned an aevanced composite
nosewheel door of the F 28 Fellowship and two sets of these doors are flying now over three years with more
than 7000 flyinghours and 7500 starts and landings with complete satisfaction. The second project was the
development of a composite speedbrake for the F 28 and will be reported in this paper. Simultaneously the
development of space vehicle components, such as solarcel panels are designed successfully.
The speedbrake as shown in fig. I is located on the tail of the fuselage and is used mainly as speed regu-
lator during the landing. This part was chosen for reason of the following characteristics:
- highly loaded hinge arms
- exposed to impact
- sonic fatigue loaded
- interchangeability
- non critical aircraft component
The a~m of this project was to gain experience in design, production and flight service with a highly loaded
component with the available technology. The following characteristics were known from the existing metal
,speedbrake design: aerodynamic performance, strength and stiffness requirements, contours and tolerances,
configurations of the junctions, quantity of impact damage. The existing Al-alloy speedbrake is based on a
multi rib construction with two forged hinge arms and Al-alloy skin sheets.
The project was divided inro three phases:
phase 1: Study of different types of structures and selection of the c~nfiguration for further design.
phase 2: Design of a schematic model of the speedbrake. Evaluation and tests of structural details. Fabrica-
tion of two schematic speedbrakes for static and dynamic load tests.
phase 3: Production of one speedbrake for flight testing on an operational aircraft.
The design chosen after phase I is shown in fig. 2. The structure is built up from the following main parts:
two hinge arms, conrtructed as rectangular CFRP beams running from the hinge points to the aft end-rib, two
honeycomb sandwich skin panels, two end-ribs, leading edge, trailing edge.
The selection was made with the following criteria: minimum weight, producibility, impact resistance, sonic
fatigue resistance, small number of parts. The main materials as used for the speedbrake are listed in tabel I.
The ribs leading and trailing edge consists of solid laminates of aramide fibre composites. The choice was
based on low stresses in these parts and the occurence of impact.A
The outside skin panel of the speedbrake is mainly loaded with a bending moment and tensile forces, and is
exposed to impact. A serie of tests was done on the impact resistance of aramide fibre sandwich honeycomb
panels in reference with the impact data of the Al-alloy speedbrake. This resulted in the optimum confi-
guration with a rwo layer aramide fabric outer face; 7,5 mm Nomex honeycomb core and a 3 layer CPpF innerface.
This combination exposed to equivalent impact loading will result in a locally damaged outer skii and honey-
comb core and an undestored inner skin. The inside skin panel is loaded in bending and in compression. It
is built up from two CFRP skins of 3 layers with a 5 mmn Nomex honeycomb core. The Nomex honeycomb was chosen
for both skins to avoid corrosion. The edges of both skin panels were partly filled with syntactic foam and
The hinge points were extensively tcoted with different designs in a static and dynamic test program with
lug type test specimens (fig. 4). The ultimate load or one lug of the speedbrake arms is 245CON. The
maximum dynamic load for one lug is 16340 N for 5.10 cycles. The first design consisted of the 8 layer ba-
sic arm laminate, reinforced on the ends with Ti.6AL.4V doublers bonded to the laminate with a 1200C curing
adhesive (fig. 4a). Ti-alloy doublers were chosen to avoid corrosion which likely will occure when Al-alloy.
is used in combination with CFRP due to electro chemical contact. Also the smaller coefficient of thermal
expansion makes the use of Ti more attractive than Al-alloys when high temperature curing adhesive are used.
The static strength at.Jstiffness were very near the calculated values and fracture occurred in the middle
of the CFRP laminate. The strength P-70550N is nearly three times the required ultimate load. The dynamic
load program was carried out with constant amplitude at a frequency of 10 Hz.
Mhe applied load range was: 2950 N - 35300 N (20% - 200% daily load), 2950 N - 26450 N (20% - 160% daily
load), 2950 N - 17650 N (20% - 110% daily load). At all load levels delamination occurred in the adhesive
layers starting at the ends of the Ti doublers and fatigue cracks started in the Ti doublers aside of the
holes. In most cases the final fracture of the test specimens was caused by the fatigue cracks in the Ti
doublers and the accompaning dglaminations aside of the hole. The test specimens loaded on the lowest load
level withstood more than 2.10 cycles without fracture and fatigue crack and delamination propagation was
very low. However the results were disappointing compared to the static test results.
In order to improve the stiffness transition at the ends of the doublers and to reduce the peak shear stresses
a setie of test specimens with tapered Ti.6AL.4V doublers were manufactured and tested with the same load
pattern. The thickness of the Ti doublers was increased with 15% and the tapering was machined with standard
milling equipment. (fig. 4b). The results of the tests showed an improved resistance at all load levels. Es-
pecially the initiations of the delaminations were delayed and the propagation was much lower. Fatigue cracks
in the Ti doublers aside of the hole again caused the final bgeakdown of the test specimeni. The test specimeas
at the lowest load level were not damaged in any way after 10 cycles at which the tests were stopped. All
test specimens were regularly inspected with the Fokker bondtester.
In comparison with the Ti reinforced lugs a test specimen was designed with CFRP doublers as shown in fig. 4c.
The doublers were also laminated of U.D. tape with a fibre orientation calculated for minimum stiffness in-
crease in the tapered part. The tapered ends were machined with diamond tippei tools. The test snecimens
were static and dynamic tested with the same load pattern as the previous lugs. The test results were more
or less equivalent with the results of the specimens with tapered Ti doublers. Damage of the test specimens
occurred both static and dynamic through delamination of the layers of the tabs near tge bond layer, star-
ting from the machined r4ges. At the loweat load level no damage was detected after 10 cycles.
Although it is likely Lnat improvements can be made with other configurations of the CFRP tabs such as, lami-
nated tapers, other fibre orientations or laplengths, this design was adopted for the hinge points of the
spaedbrake. The main advantages of this design are: low weight, lack of thermal stresses and elamunation of
corrosion problems. Also it was a challenge to test such a configuration in service.
After the evaluation of the results of the lug type specimens it was decided to design two different confi-
gurations for the actuator attachment point. Otte with Ti side plates (type I) and one with CFRP side plates.
(type II). The test specimens are shown in fig. 5. The configuration chosen for the tests makes it possible
to test two actuating points at the same time and is also a try-out for the production process. The confi-
guration with Ti plates differ in the way, that no U section was used between the Ti places, in which the
forces were introduced. All other joints were bonded with a 1200 C curing adhesive without the assistance
of mechanical fasteners. Both the Ti-6AI-4V and CFRP plates were tapered along the edges to achieve a uni-
form shear stress in the bondlayers.
One test specimen of each type was loaded until fracture which occurred for the type I specimen at 81910 N
and was caused by shear failure in the bond layer and in the outer layer of the CFRP beam laminate. The
type II test specimen failed at 84860 N due to shear in the layers of the CFRP plates near the adhesive layer.
The requLsted ultimate load for the speedbrake actuator point is 39600 N which is far below the tested strength.
The dynamic load for the speedbrake is equal to limit logd (26400 N). The type I test specimen was loaded
between 490-29430 N with a frequency of 2 Hz. After 3.10 cycles some small delaminations were found in the
bond layer between the tapered edges of the Ti-plates and the CFRP beam. During the rest of the test no crack
propagation was found. The test specimens were regularly inspected with different non destructive inspection
techniques such as: the Fokker Bondtester, Ultrasonic inspection, X-ray and one time with neutrography.
"After 5.10 cycles a change of response with the bondtester was found in the middle of the bonded area of the
Ti plates. This area grew very slowly during the test and is expected to be a degradation of Lhe adhegive
layer or the CFRr laminate due to small delaminations. The dynamic load test was stopped after 2.5.10 cycles
9 and a residual strength test was carried out. This resulted in fracture at 70630 N about 15% lower than the
initial static strength.
The 5ype II test specimens with CFRP plates were cycled on tne same load levels as the type I specimens. After
5.10 cycles no damage was found. Therefore it was decided to increase the load level to 490-39240 N. A; 10
cycles some smal! delaminations were found in thq bond layer between the 01FR' ates and the beam near the
lug. The different indication with the Fokker bondtester in the middle of the bonded area occurred similar
to the type I specimen. The test was continued with a further increased load level of 490-49000 N. This re-
sulted in a fgst propagation of the delaminations and no measurable enlargment of the questionable area.
:. After 1.05.10 cycles the test was stopped just before total fracture.
Although both configurations were effective for the speedbrake the design with CFRP plates was chosen for
the ac.uating points for the same reason as the hinge points. (fig. 6).
The main assembly of the speedbrake structure is bonded with a roomtemperature curing adhesive without the
use of additional fasteners. The critical joint between the arms and the skin panel is loaded in flatwise
tensile. A test program was carried out with different adhesives and configurations to prove the durability
of this joint. The test results indicated a sufficient strength of the arm to skin joint for the speedbrake
design.
"After completingthe details tests, two schematic models of the speedbrake were built. (fig. 7). From the
point of view of strength, dimensions and production this model is equal to the real speedbrake. The main
,, problem during the assembly was to achieve an uniform thickness of the bond layers between the arms and the
skin panels. The mismatches results from the use of positive laminating moulds for the beams and nTegative
moulds for the skinb. The models were mounted in a test rig representing the fixation to the aircraft and
the aerodynamic forces on the outside skin. The first model was loaded until fracture and broke down at
1.75 times ultimate load. Fracture was initiated through bending of the arms. The second model Was dynamic
loaded between 1,5% - 65% ultimate load at a frequency of 0,5 Hz. The test was stopped after 10 cycles
without any significant damage. A very small delamination was found in the adhesive between the arm
t
7-3j
and the outer skins near the front rib. The design if the speedbrake is adapted by the extension of the outer
skin panel with two tapered lips in order to decrease the high shear stresses in the adhesive layers. The
good results of the tests on the model speedbrakes made it possible to copy the design with only small im-
provements for the operational speedbrake. Most of the difficulties when building this part, originated in
the complicated three dimensional curves and in the control of the processes and tolerance3.
In October 1978 the third phase was finished with the installation of a composite speedbrake t0 the Fokker
F 28 Fellowship of the NLM. It is flying now over a year to the complete satisfaction of the operator, with
1500 flying hours and 2000 starts and landings. Apart from the normal schedule the speedbrake is checked
regularly following a special inspection scheme on the aircraft and in our inspection laboratory. No damage
due to normal service has been found until now after 1500 flying hours. The weight saving for this part was
about 25% compared with the Al-alloy speedbrake.
0
CFRP F & H CARBOFORM CODE 92; UNDIRECTIONAL PREPREG TAPE 120 C CURING EPOXY;
AFRP NARMCO 550; 181 SATIN HAMNESS; 120C CURING EPOXY
0
ADHESIVES BSL FM 123-5; 120 C CURING EPOXY 3M EC 2216A/B; ROOM TEMPERATURE CURING EPOXY
HONEYCOMB NOMEX
fi
V
i4
, -
7-4
3960 N
-~ I
('ID
2S 10
'4A
1fl YE1
-1:
HIG AR
'A ~A '4
! -|
SUMMARY
The present work describes investigations to obtain generic information on the time-
dependent environmental behavior of graphite-epoxy composites used in aerospace structural
applications. Experimental and analytical results are presented for the individual and
coupled effects of temperature and absorbed
tional and cross-plied composites. The scopemoisture
of the on the creep compliance
investigation of unidirec-
includes measurement
of tensile creep response at room temperature and four elevated tevreratures (up to the
glass transition region of the materials) for dry specimens and for spezimens moisture-
saturated at 75% and 95. relative humidity. Data are presented for two material systems:
Narmco T300/5208 and Hercules AS/3502. The stress levels used are demonstrated to be
4 within the range of linearly viscoelastic response. Superposition techniques are used to
determine compliance master curves from the results for individual temperature and mois-
ture conditions.
INTRODUCTION
The use of graphite-epoxy composites for aerospace structural. applications has evolved
as a result of their mechanical properties, such as high strengt,-to-weight ratio, and
their tailorability of design, relative to conventional materia'.. The composite consti-
tuent materials are usually assumed to respond to external inruts in a linearly elastic
manner, and this feature is reflected in the common analytical methodologies used. It
must be recognized, however, that composites do exhibit a significant amount of time, tem-
perature, and moisture dependent behavior, particularly in service-type environments. In
fact, the problem of time influences the behavior of these materials in several ways; e.g.,
the tim, of'exposure to environment, the age of the material, and the time (duration) of
a particular loading condition.
Although much research has been conducted on the viscoelastic behavior of polymers
and particulate-filled composites, only limited investigations on the time-dependence of
fibrous composites had been accomplished through 1975 (Ref. 1-7, e.g.). Recent years
have witnessed rapidly expanding activities directed more specifically at time-dependent
behavior of advanced composite materials systems of particular interest for aerospace
applications (Ref. 8-11). These iorks have established that although the behavior of
fibrous
range of composites
input is in general
(temperature, not linearly
humidity, viscoelastic, there ismayoften a considerable
appropriate approximation. load) where linear response at least be a very
In the present program, we have undertaken a rather basic attempt to obtain generic
information on the time-dependent mechanical response of commonly-used graphite-epoxy
materials and to analyze what that time-dependence means within the framework of existing
methodologies. We have obtained a fairly complete linearly viscoelastic characterization
of the principal properties for two material systems over a range of temperature and humid-
ity codditions representative of (or exceeding) service conditions. We have then used
this experimental characterization with conventiona1 analytical techniques to predict
time-dependent laminate response, in order to ascertain the magnitude and importance of
time, temperature and humidity effects on structural composites.
The overall features of the test/prediction scheme are indicated in Fig. 1. The six
types of coupons shown were tested for tensile creep as a function of time (t), tempera-
ture (T), and humidity (H). From the (0)6, (90)15, and (45) . results, the principal
moduli were extracted. These moduli were then used to predic? the measured longitudinal
stiffness strength (oR), and creep tests for specimens with non-uniform moisture distri-
button (M(z)) are also included in our test program, but are beyond the scope of the
present discussion.
EXPERIMENTAL CONDITIONS
The present program has included tests on two material systems, Narmco T300/5208 and
Hercules AS/3502. Although the fibers are different, the matrix materials for these 4500 K
(350 0 F) curing materials are quite similar, consisting in each case of a major and minor
8-2
TESTS: t, T, H,M(z),o'R(RT)
E1l1(TH)
PROPERTIES PIMA PREDICT Ex(tTHM(z))
E22(t.T.H)
G12(t,T,H)
Longitudinal and transverse expansional strains due to temperature and moisture were
measured under various coupled and uncoupled conditions of temperature and humidity.
Thermal expansion measurements on both dry and wet specimens were conducted with a Perkin-
Elmer TMS-l Thermomechanical Analyzer (THA), with temperature and sample dilation monitored
and recorded with a Hewlett-Packard 3050B data acquisition system. Moisture-induced swell-
ing measurements were accomplished with four-place micrometers, using alignment fixtures
to insure repeatability of specimen alignment and reference point position. Typical trans-
verse expansion values on unidirectional specimens (both materials) with respect to tem-
perature and percent moisture content by weight were a 2 2 = 25gd/K and P22 = 3000 .E/70M.
Figure 2 shows representative swelling data represented by percent change in width
(transverse to fibers) as a function of moisture content. We note the usual threshold
value of moisture required before swelling begins ( 0.11. in this case), the similarity of
behavior for the two materials, and the apparent temperature/moisture coupling caused by
elevated temperature (ET) conditioning (322 and 339 0 K) as opposed to room temperature (RT)
conditions.
0.5 I I- -T
CA -- T300/5208 0o- -
lAS/3502 I o
0.3 -- - I
02j
3
&W=0.28(M_0.I)l. (RIT)
The test matrix for mechanical testing used in these investigations is indicated in
A . Figure 3. The dashed curve in the figure shows our measured dependence of the glass tran-
sition temperature (Tg) on moisture content. Using room temperature and the Tg as the
S.temperature range of our studies, we conducted time-dependent (primarily tensile creep)
tests under the conditions indicated by the squares, and baseline stress-strain tests
where the X's appear in the figure. The baseline stress-strain to failure tests were con-
ducted prior to any time-dependent tests in order to assess the trends of mechanical re-
sponse changes to be expected and to aid in selecting stress (strain) levels for creep
(stress relaxation) tests.
.'- ! All creep tests were conducted on sets of three replicate specimens in deadweight
( 4
SI' -creep machines. Several rapid load-unload cycles were used for mechanical "break-in" of
the coupons prior to the first creep cycle. In general, the same set of specimens was
used for all five test temperatures at each dry or wet test condition.
=711Mim
8-3
10.
DRY 5020()0
[3 CI i
76 R 1.0
I D CREEP
95% RH X STATIC 1.5
Filure3 TestMatrix
Dry specimens were desiccated under vacuum/temperature (339 0 K) for at least seven
days prior to test, and desiccant was maintained in the oven during dry cieep tests.
"Wet" specimens were preconditioned to tmiform moisture content at either 75% or 957.
relative humidity (and 339 0 K) prior to test. During the creep testing, the wet specimens
were immersed in appropriate mixtures of ethylene glycol and water to maintain the mois-
ture content attained during their prior conditioning. For these wet tests, the strain
gages were coated with Micromeasurements M-Coat to retard (but of course not prevent)
moisture effects on the gage adhesives. S.rain gages would typically function without
problems for about five days, which was more than adequate for testing at 5 temperatures.
Essentially all creep runs were 100 minutes in duration, followed by a recovery (no
load) per.od of 100-300 minutes, depending on the next test temperature. Stress levels
for creep testing were set at typically 25-357. of ultimate values for each temperature/
moisture condition. Verification of linear response was obtained by superposition of
creep-recovery results, as shown for a typical case in Figure 4.
-.. - - 1, I
- 8.13 _.
1.uu
L7.88
S7.63 121
ca
TIME (MIN)
Figure 4 Linear Viscoeigsticity of T300/5208 (90)15: 95%RH, 150OF
(primThe overall plan of testing and analysis has stressed (1) time-dependent measurements
(primarily 100-minute creep) in the region of linear response; (2) generation of oompliance
master curves by simple horizontal shifting of data (time/temperature/humidity superposi-
tion); (3) extraction of moduii using quasi-elastic conversion; (4) prediction of lami-
nate response using principal moduli master curve data points to compute averag- laminate
stiffnesses. We will discuss specific results related to the first three items above in
this section, and example predictions will be presented in the following section.
P!1
8-4
Uj T3,H2
W /
T1,H1
TIME (MINUTES)
nF uFigure 6 Typical Ella nd v12 For FT Wet (0J6 Coupons
In Figure 7 are summarized our data for tensile creep compliance of dry (45) cou-
pons. The data points for the T300/5208 system are horizontally shifted log-log results
(referenced to RT) from 100 minute creep tests at 6 temperatures (an extra test was per-
formed in this case at 464 0 K (375 0 F) somewhat above the dry Tg). Several features are
4. noteworthy in this figure. (1) A smooth time-temperature curve may be derived from simple
horizontal shifting. (2) The two materials have similar mechanical behavior, as evidenced
by .he dashed AS/3502 master curve ZoMoared to the T300/5208 data points. (3) Longer-
term tests agree rather well with sho-t-time tests, as indicated by the elevated tempera-
ture 5000 and 7000 minute results. (4) The shift factors, aT, shown in the inset are
'1 essentially identical for the two materials.
Using this compliance master curve, we have produced the Gl2 (t) curves shown in
Figure 8. The tensile modulus ER has been taken euqal to (Sxx)- , and the shear modulus
"isobtained from ER(t) and the measured (time-insensitive) Poisson ratio vx1 for this
if coupon. The G1 2 (t) curves are given in Figure 8 for various reference conditions as in-
dicated. With respect to RT dry conditions, time scales of 15 years present minimal
48-
concern, for the modulus changes by only about 20%. As the material becomes warm and wet,
however, comparable decreases in modulus may be expected in times of the order of weeks.
S~.09
-825 0 0 T300/5208 (':46)2 DRY 0
-1 0 AS/3502 ,,0 R
-2, T300/5208
100 MINUTE TESTS C-
-6.39 -3
7" -4 C, .07 k.
.4 3/3502 o I
9-6.35 -5 100 MINUTE TESTS
" S-7 ~I/T(0K)X '103 " ,z OO , .6 , N.i
TESTA''
X -6.0-. 2.2 2.4 2.6 2.8 3.0_3)40 3000 F
-6.45 ;;-7-l . 1;K---A5000 MIN15o
TEST .0 Ix
0 1 2 3 4 5 6 7 8 9 1.
LOG t/#T (MIN)
Figure7 Crep Compliane Mater Curve for Dry (+45)
1.25 Ie
.2 WET (75%RH 4.
, -1
-8 225VF OR -4.0 r
1.10 "J-2- RT WET
The inset to Fig. 8 presents the log aTH shift surface for a range of t-mperature
and humidity conditions. We note that similar shifts accure for RT/95% RH and for 370K/
dry conditions. When both moisture and elevated temperature are present, however, the
shifts are considerably greater.
Our results for the time-depcndence of E22 are generally similar to those presented
here for G12, although the transverse modulus exhibits slightly less relaxation than does
G12, a result similar to that of Ref. 8. We have also generally noted considerably more
scatter in our measurements on 900 coupons than Is the case for the (45) specimens. We
investigated the possibility of damage in the 900 coupons by using replicate, techniques
to search for edge cracks. No evidence of damage was found 'or any of the 3pecimens
examined.
Figure 9 demonstrates our results for the 900 coupon compliance, S22, for several
possible cases. The shaded curve represents our derived master curve for dry T300/5208
coupons, all of which were approximately one-year old (time since cure)- If we superpose
our wet" results, using our previously-derived shift factors from the G1 2 case (assuming
all time-dependence ii. due to the matrix material), then the solid curve results.
The apparent discrepancy bet.,een wet and dry results at long-reduced time is obvious.
Two possibilities immediately ario. (1) The effects due to temperature and moisture are
not superposable in the same manaer, although this assumption yielded reasonable results
in the Gl2 case. (2) The dry speci. ns have "aged" (see e.g., Ref. 13) to a greater ex-
tent, yielding a lower compliance v- long times.
were then shifted by a log aT previously derived from G12 data at that temperature. The
results are indicated by the data points in Figure 9.
L ON DRY, "UNAGED"
-6.10O
COUPONE
C" -6.15
RANGE OF
-6.20 RESULTS FOR
DRY "AGED" (-1 year) .09
COUPONS
4-.25 L L L I I
5 2 4 6 8 10 12
LOG 4 (MINUTES)
Figure 9 S22 Master Curves for T300/5208
Although between
the agreement these latter results must still be
the wet results and the dry regarded as preliminary at this time,
results for "fresh" specimens is strik-
ing.
PREDICTIONS
Using principal moduli derived from our compliance curves using quasi-elastic conver-
sion, we have performed some preliminary predictions of expected laminate behavior. The
thrust of our work to date has been to see how well we can do using familiar analytical
techniques, so our first predictive attempts have concentrated on using time-dependent
data in conventional elastic analyses.
As an example, Figure 10 shows Halpin-Tsai predictions for Em, the effective matrix
modulus, P-nd for E22, based on 6ur experimentally derived G1 2 master curve (wet aud dry
results). We have used fiber properties essentially the same as those given in Ref. 14.
The data points (o) are our experimental data f'zm this program while the solid curves
are Halpin-Tsai results. The vertical bars on th3 E data represent typical scatter in
our 900 coupon results. The data points (x) are results from Ref. 15 on neat 3502 resin,
which is similar to 5208, as we have noted. The data points (A) are Aerived from results
in Ref. 16 for 5208 resin.
r E22
-HALPIN-TSAI CALCULATION
1.0
0.9
L1 G1
122
01
0. IA _ -4_
0O 0T300/5208 (PresentWork)
3502 RESIN (Ref. 15) Em -
05200 RESIN (Ref 16)
00 2 4 6 8 10
LOG 4 (MINUTES)
Figure 10 Measured and Predicted Laminate Properties
Several features are worth noting from Figure 10. If we base all calculations on
-V G1 2 results, i.e., if we assume all time dependence is due to the resin material and is
-roperly described by our measurements on ( 45) tensile coupons, then using reasonable
*,'lues for fiber properties and geometrical parameters, with no unusual adjustments of
~ "parameters, we obtain a very reasonable description of E2 2 results. Additionally, the
predicted matrix modulus agrees very well with independent measurements on the cured resin
material.
T
8-7
Given then the time-dependence for G1 2 and E 2 2 shown in Figure 10, we have used an
available elastic code to calculate the average stiffness, as a function of time, of a
(90/45). laminate. For these calculations we used Ell - 131 GN/m 2 (19 MSI), and V 1 2 =
0.31, which are "average" results from our weasurements. Both of these latter moduli
are assumed insensitive to time.
Figure 11 presents the prediction obtained on this basis, along with the experiment-
ally-obtained master-curve data for the (90/45)s T300/5208 laminate. The agreement be-
tween measurement and prediction is within about 47. over the entire range of reduced time.
6.58
LAMINATE PREDICTION 26
FROM MEASURED
6.56- PRINCIPALMODULI 25
6.64 -24
6.5 T300/5203 2
I (9U/:45)s I
2 23 2
6.4 21
0 2 .4' L 10
l LOG 4IMINUTES)
CONCLUSIONS
The following statements summarize the main conclusions from our work thus far:
2. Ell and u12 are not sensitive to time, temperature and moisture.
position.
S3.
Long-term iesults are predictable from short-term tests by simple horizontal super-
S4. Prediction of laminate master curves is possible using conventional analysis procedures.
10. W. J. Renton and T. Ho, "The Effect of Environment on the Mechanical Behavior of
AS/3501-6 Graphite/Epoxy Material," ATC Report No. B-92100/8CR-105, Vought Corpora-
tion, Dallas, TX, August 1978.
12. J. Romanko and W. G. Knauss, "On the Time Dependence of the Poisson's Ratio of a
Conmnerci~al Adhesive Mate-ial," J. Adhesion, 10 (to be published), 1979.
13. L. C. E. Struik, Physical Aging in Amorphous Polymers and Other Materials, Elsevier
Publishing Co., New York, 1978.
16. K. G. Kibler and H. G. Carter, "Viscoelastic Parameters of Epoxy Resin from Thermo-
mechanical and Electrical Conductivity Measurements," Composite Materials: Testing
and Design (5th Conference), ASTM STP 674, S. W. Tsai, ed., American Society for
Testing and Materials, 1979, pp. 282-288.
ACKNOWLEDGEMENTS
This work was accomplished under Contract F33615-77-C-5109, sponsored by the Air
Force Materials Laboratory (AFSC), Wright-Patterson AFB, OH 45433. Dr, J, M. Whitney
(AFML/MBM) is the Air Force Project Engineer. Drs, R, A, Schapery and Y. Weitsman, of
{
-A Iprogram
Texas A&M University, and Dr. W. G. Knauss, of California Institute of Technology, are
associates. Other participants at General Dynamics deserving special acknowledge-
ment include T. Crjasy, J. Fruit, J. Reynolds, W. Robinson and J. Halkias.
X
4
R2-1
RECORDER'S REPORT
by
Dr G.Dorey
Materials Department
Royal Aircraft Establishment
Farnborough, Hampshire GU 14 6TD
England
Information is needed on the effects of simultaneous application of load, temperature and humidity on composite
materials, to generate confidence that they will perform adequately in the service environment. Reliable analytical
methods of predicting performance will obviat,, nuch tedious testing. These four papers provided useful additional
data. Three of the papers were on fatigue, one on creep compliance; three were on coupon testing, one on an aircraft
component.
The paper by Arendts et al on the fatigue of (0, 45, 90) laminates concentrated on compression, to load the
matrix, in flight-by-flight testing at room temperature, with five different loading spectra to represent different aircraft
components. Miner's rule was found to be unsatisfactory because composites appear to be particularly sensitive to peak
loads, overloads causing a significant decrease in fatigue life.
Gerharz and Schfitz tested (45) laminates, again to load the matrix. Flight-by-flight fatigue tests, so far only at
room temaperature, showed that degradation in mecha-:cal properties could be related to increased damage shown by
ultrasonic C-scan. Static tests showed that elevated temperature and moisture significantly reduced strength and modulus,
and this could well affect the fatigue results when environmental fatigue testing is done. The paper contained a useful
literature survey.
K bKibler's paper also showed that temperature and moisture has a marked effect on the mechanical response of matrix
dominated properties. Measurements on simple (0), (0, 90) and (45) laminates can be used to predict the behaviour of
; more complex combinations of these orientations, and 100 minute cicep tests can be used to produce master creep
curves by a simple superposition procedure.
e
. Dijns gave useful information on an actual component, an F28 speed brake. Careful testing of structural elements
S .'- and load attachment points showed the superior fatigue properties of CFRP compared with titanium and aluminium
alloys, and led to an all-composite construction incorporating aramid :ibres for impact resistance. !n service, the com-
} , If ponent showed no detectable damage after 2000 landings.
The discussion concentrated on whether laboratory testing could adequately represent the service coriitions.
, l There was concern about the different specimen designs used by different organizations, and about problems with
moisture gradients produced by accelerated testing.
SI"The session emphasized that effects of load, temperature and humidity have a most significant effect on fatigue
and creep when the loading is such as to produce !-gh stresses in the matrix, such as compression, shear and out-of-plane
stresses. Combined effects of moisture and temperature are significantly worse than either alone. But if fibres are
aligned in the direction of the stresses the effects are much smaller. There seems to be no acceptable way at present of
predicting fatigue lives and representative flight-by-flight testing is necessary, but more data will help in producing
I[ ".analytical models.
I 'lthough degradation due to fatigue, temperature and moisture can be demonstrated in the laboratory, the data
available suggest that, with present design limits, safety factors are adequate and there is no significant fatigue problem.
Indeed the practical component described in this session has performed well in service and this was confirmed by
,experience with other components in service described in Session V.
1j.4
9-.
by
G Dorey
Materials Department
Royal Aircraft Establishment
Farnborough, Hampshire, GU14 6TD, England
SUMMARY
A variety of CFRP laminates and one GRP laminate were subjected to impact by steel balls, over a
range of incident energies, and residual strengths were measured; superimposed static load during impact
substantially altered the residual strength curves. The laminates were also tested with machined notches
and analysed in terms of fracture toughness; results were compared with impact performance both during
impact and in subsequent residual strength tests, with the aim of correlating service performance with
laboratory toughness tests. Effects of materials and geometric variables are discussed together with
possibilities for improvements.
I INTRODUCTION
Advanced composite materials, being introduced into aircraft structures because of their high specific
strength and stiffness, must perform adequately in the service environment. One possible threat to their
performance is that of damage from rough handling or from impact by foreign objects. At low energies this
can result in surface damage and delamination and, at higher energies can produce through-thickness cracks
and holesi. The damaged area may be further affected by temperature, humidtty and fatigue loading, but one
of the most damaging effects is that of broken fibres reducing the stall ic tensile arid compressive strengths.
Full-scale testing of composite components is expensive and time consuming and bas to be kept to a
minimum. Useful indications of the impact performance of composites can be obtained from coupon tests which
simulate service threats. However, this still involves relatively large numbers of specimens, to allow for
a range of incident energies, various geometries and d-fferent residual mechanical properties. Simple
meclianical tests are required that correlate with impact performance so that materials parameters can be
modified, improvements made and the most promising materials selected for n'ore searching investigations.
This paper describes testb on a variety of CERP laminates made from several fibre-resin systems using
several different lay-ups. Foreign object impact was simulated using dropped weights and steel balls fired
Residual
tests.strengths
from an airgun,insome
were measured or coupons
of the
flexure tension being stressed
and the results tocompared
uimulatewith thoseof from
impact loaded panels.
notched tensile
2 THEORETICAL CONSIDERATIONS
*ii In aszessirg the imract performance of composite laminates consideration has to be given to both the
processes occurring daring impact to produce damage and the effect of this damage on residial strengths.
2.1 Impact
The damage done during impact depends on how the stresses build up, on how they compare with the
various materials strengths, on how they are redistributed as the fracture progresses, on the fracture
energies and on the energy available to drive the fracture processes. These in turn depend on the material
properties and geometry of the test specimen and on the energy and momentim of the projectile. No damage
need be done if the energy of the projectile can be accommndatd oy elastic strain energy in the specimen.
1
Simple calculations have been made of energies necessary to cause:
3
a. shear fractur- /E)(w1 /t)
_ __ __ __
-I
9-2
When the local stress at ihe tip of a notch reaches the strength of the laminate, either the
laminate fails or local crack blunting processes occur, such as shear cracking, which reduce the stress
concentration. In some laminates such as (0) or (+45), these shear cracks can remove the stress
concentrating action of the notch and failure does not occur until the net section stress equals the
unnotched failure stress. But in (0,90), (0, 90 + 45) and (0, 45) CFRP laminates, failure can occur
when relatively small volumes of material at the notch tip are at or near the failure stress, while the
average net section stress is still significantly less than the failure stress.
Bishop and McLaughlin4 used a notch blunting mechanism to explain their results on (0 + 45) CFRP
laminates; with increasing stress,shear cracks at the notch tip increase its effective radius of curvature
until a critical radius of curvature is reached, characteristic of the laminate, when failure occurs. The
critical applied stress ac is given by:
Oc +o/(i
q (a/po)) (1)
where To is the strength of the unnotched laminate, 2a is the notch length, q depends bn the elastic
anisotropy and po is the characteristic radius of curvature; typical values for p 0 for CFRP laminates
were about 10 mm for a tough laminate and about 5 mm for a brittle laminate.
5
Potter invoked a similar crack blunting mechanism but he argued that, as the stress and hence the
notch tip radius both increased, the stress gradient would decrease, and failure would occur when the
stress gradient was reduced to a critical value characteristic of the material. Simplifying the mathematics
for relatively sharp notches, the critical applied stress is given by:
where C depends on the elastic anisotropy and S has the dimensions of length and contains terms that
depend on material and laminate properties.
Nuismer and Wnitney6 used an "average stress criterion" in which they did not specify the damage
at the notch tip but considered that failure occurs when the elastic stress, averaged over a characteristic
distance a ahead of the notch, reaches the unnotched laminate strength. This gives:
For multidirectional CFRP laminates a0 is typically 2 to 4 mm; the model has also been fitted
to notched compression data and a larger value of ao was found, namely 6.2 mm.
Fracture mechanics has been applied with reasonable success to notched multidirectional laminates236
In this case the shear cracks at the notch tip are considered not so much to blunt the notch but rather to
form a damage zone, similar -o the plastic zone in a metal, that effectively lengthens the rotch and increases
the release of strain energy. Thus the fracture toughness K becomes (r, (n(a + rp))I- and
models of equations 1 and 2 and betweenr the stress-integrating models of equations 3 and 4. Although all
four models use different physical mechanisms, they all depend on the form of'the stress distributicn ahead
of the notch and it is not surprising therefore that they all have a 1/4-a dependency. All four equations
can be fittid to most notched tensile strength data on multidirectional CFRP laminates, if the two materials
constants in each equation are suitably adjusted. The models associated with equations I to 3 have the
advantage of being applicable to notches of various shapes, whereas equation 4 depends on laminate toughness
and is not related to the unnotched laminate strength. Although equation 4 applies strictly only to sharp
notches, Avery and Porterd have found that it can be used to predict residual strengths of multidirectional
I CFRP laminates with irregular impact damage if 2a. is put equal to the "transverse lateral damage" (TLD)
I and a larger value of K' is used than that for sharp notches. All four models a-ae c c_.e form of stress
xedistribution by the formation of a damage zone at the notch tip and it iEs not unreasonable to assuime
that the size of the damage zone will depend on material and laminate properties, especielly on shear
strengths and shear fracture energies, that control shear cracking. Increased damage zone size increases
the amount of work Gc needed to propagate the crack and therefore, since KX2 = GO E, larger damage zones
tend to be associated with tougher laminates, although if the energy density within the damage zone could be
increased, toughness could be increased with less associated damage. The formation of the damage zone also
'IsZ
9-3
increases the volume of highly stressed material at the notch tip and therefore local variations in
material strength, such as variations in fibre strength, become important in considerations of the
statistics of failure probability.
3 EXPERIMENTAL DETAILS
All the laminates were made from preimpregnated warp sheet, layed up in the required orientations
and moulded in an autoclaye to give balanced symmetrical laminates with 60% by volume of fibres; a summary
of the laminates ia given in C'able 1. Specimens were cut from the laminates using a diamond slitting
wheel, and for tensile specimens aluminium alloy end tabs were glued on with an epoxy adhesive. For notched
specimens, saw cuts were made witha 0.25 mm jewellerst saw, or circular holes were drilled with diamond-
tipped drills.
Charpy impact specimens, 50 mm x 10 mm and unnotched, were cut from the multidirectional laminates
and tested in a pendulum impact machine fitted with a pendulum of capacity 7.5 J; the span was 40 mm and
the specimens were mounted so that the deflection was out of plane.
For drop-weight impact, a 12.7 mm diameter steel ball was loaded with weights to adjust the incident
energy and dropped through 1 m on to plain rectangular specimens, simply supported over a span of 20 mm
for subsequent interlaminar shear testing or a span of 50 mm for subsequent flexural or tensile testing.
Ballistic impact was by 6 mm diameter steel balls fired from an airgun, the air pressure in the cylinder
being used to control the ball velocity and hence the incident energy; again the specimens were plain
rectangular and simply supported'on parallel supports 50 mm apart. Specimens were 25 mm wide for laminates
4-6 and 50 mmwide for 'laminates 7-13 and the GRP. For ballgun impact on loaded laminates, plain rectangular
specimens 30 mm wide were mounted in a simple screw-loading frame and the load was measured by calibrated
strain gauges on the frame arms.
Fig I shows the effect of machined centre notches on the tensile strength of laminate 1. It can be
seen that the notches reduced the tensile strength by more than the loss in cross sectional area; ie the
laminate was notch sensitive. The tensile strength was reduced more by the sharp notches than by the
circular holesk. The data for the sharp notches are replotted in Fig 2 as fracture toughness,
KC= ca 0 c(it(a + rp
p))i, together with data for laminates 2 and 3, where a is a finite width correction and
r is the damage zone correction mentioned in section 2. The values of rp, chosen to give the best fit
p
for the data, were 0.2 mm for laminatd 1, 1.2 mm for lam.inate 2 and 1.6 ms for laminate 3. Observations
in a scanning electron microscope of the fracture surfaceas near the notch tip showed a surface roughness
of 2-3 mm for laminate 3 and 0.4-0.5 ,mafor laminate '. Thus fracture toughness was found to predict
reasonably well the residual strength of notched laminates with K between 20 and 50 VIm-3/2, for notches
from 2 mm to over 40 mm in length. For notches longer than about ten times the damage zone size (a > 5 r
the correction became insignificant compared with the experimental variation in a',
were the normalized net section stresses for a 8 mm centre notch in a 25 mm wide specimen, and this data was
used to calculate the fracture toughness shown. It can be seen that although the unnotched strengths were
similar (variations of 30% for flexural strength and 25% for tensile strength) the notched strengths varied
by a factor of 4.5 and the Charpy impact energies by a factor of 7.5. In Fig 3 the Charpy impact energy
is plotted against the fracture to.Thness for laminates 4 to 6 and for six other laminates (three other
fibre/resin systems with the same a and b lay-ups), showing a good correlation. If the Charpy impact'
A energy is related to the fracture energ,- G then, beca se K
2 = G E, one would expect a curve of this form.
This would only be true if the stored enert.r at the point of failure in the Charpy test, given in section 2.1
and in this case about 0.3 J, is less than tA.e total fracture energy ywt, in this case about 1.2 J. For
all cases except laminates 6a and 6b, lay-up a, with the 450 layers on the outside, was both tougher than
lay-up b and gave greater impact energies. The tzrpes of fracture seen in the Charpy impact test are
indicated the tcugher laminates showed a fibrous fracture whereas the brittle laminates 6a and 6b
in Fig 3;clean
showed relatively fractures. Fig 4 shows the dqcrease in interlaminar shear strength caused by drop-
weight impact using a small span/depth ratio that resu.ed in delamination. Figs 5 and 6 show the reductions
in flexural strength caused by flexural failures ir the crop-weight and ballgun tests. In all these tests
-m" resin C (laminates 4a and 4b) gave tougher laminates than resin B (5a and 5b) and resin D gave paticularly
brittle laminates. Resin D had a large relative density, 1.48 coinparel with typical values of 1.25, and a
large Young's modulus, that gave laminates with good compressive properties but poor notched tensile
properties. Lay-up (a) was generally superior to lay-up (b); reference to section 2.1 shows that this
was because, although the strengths were similar, having the low modulus 450 layers on the outside caused
2
a lower value of flexural modulus, a lower effective E ang therefore a greater va~ue of- /E. It is also
worth noting that in the case of blilgun impact, having 0 layers on the outside caused a marked minimum
in the residual strengh cur-ve and this coincided with the maximum amomnt of transverse cracking and occurred
approximately at the incident energ; needed to cause penetrution of the laminate.
9-4
Laminates 7 to 13 did not cover such a wide range in toughness as laminates 4 to 6, but were more
recent materials with greater strengths and greater toughnesses as can be se.,n from Table 3. For comparison
an E-glass laminate was included. The values of anko and fracture toughness were calculated from tensile
specimens 50 mm wide with centre notches 15 mm long (for tougher laminates larger values of w are
needed so as to keep an*. well below unity and hence give a meaningful Kc). It can be seen that, although
the value of fracture toughness of the E-glass lminate was similar to those of the carbon fibre laminates,
the Charpy impact energy was considerably greater. This was because the E-glass fibres have a Young's
modulus only a third that of the carbon fibres; dividing the Charpy impact energy for GRP by three makes
the value for GRP similar to those for CFRP Fig 7 shows that there was an increase in Charpy impact energy
with increase in K , similar to that. shown for the (0 + 45) laminates. Figs 8 and 9 show residual strength
curves for drop-weight and ballgun impact. There was not much difference between the laminates made from
the two resins, resin E being only slightly better than resin B. There was more effect from the fibres.
The qpe 3 fibre laminates (fibre 2) were weaker in the unnotched strength tests, had lower values of
fracture toughness and were generally worse in the impact tests. Laminates made from fibres 1 and 3 were
the toughest and performed best in the impact tests. Laminates made from fibre 4, although no weaker in
the unnotched strength tests were significantly less tough than those made from fibres 1 and 3. In
particular laminate 9 had a large flexural strength undamaged but showed a marked reduction in flexural
strength for drop-weight incident energies greater than 2.5 J. There was no significant decrease in the
strength of the GRP over the range of incident eaergies studied; the only damage in the GRP wag delamination.
Figs 10 and 1! show the effect of superimposed static load during ballgun impact on the residual
tensile strength after impact. For static loads up to 35% of the ultimate tensile strength (UTS), the impact
of the ball produced damage, and hence reduction in strength, simlar to that for unloaded specimens. For
static loads of 40% UTS the specimen failed on impact; similar effects have been observed in compression 7 .
In Fig 11 the dashed line is for residual tensile strengths equal to the applied stress; one would therefore
not
The expect experimental non-zero joints below this line (tensile stress greater than residual strength).
three points just below the ne resulted because of the slight relaxation of static stress in the
loading frame on impact. If there had been no effect of static stress up to the residual strength one would
have expected the sudden drop in residual strength for static stresses of approximately 50% UTS; but the
drop in residual strength for applied stresses of 30-38% UTS indicated that the applied static stress caused
a slight increase in the extent of the impact damage, that is an increase in the number of 00 load carrying
fibres broken.
5 DISCUSSION
The residual strength of a notched or damaged laminate can be predicted from the laminate fracture
toughness. The fracture toughness Kc depends on the fracture energy Gc and the effective Young's modulus E,
2c
K = GcE. Thus for CFRP laminates with similar values of E, the toughness depends on the fracture energy
c hence
and c the size and energy density of the damage zone.
The tou;hness can be increased by reducing the
fibre-resin bond strength to encourage shear cracking and by increasing the thickness of the layers to
promote longer shear cracks in each layer418. For the GRP laminate the value of Kc, and hence the failure
stress a for a given crack length, was similar to that of the CFRP laminates. Because of
the lower value
j]of E, a2/E would have been greater for GRP and hence the strain energy release rate must have been greater
as the fracture progressed, but this must have been counteracted by the increased energy needed to produce
a larger damage zone and to break the glass fibres within the damage zone. To increase further the fracture
toughness of CFRP laminates either G must be increased or the failure mode must be changed, as in the case
of 0 composites and +45 laminates, where extensive shear cracking can effectively remove the stress
concentration. But such shear cracking may be undesirable because of its effect on other properties such
as fatigue .3trength or environmental degradation.
Although the toughness of the composite laminates was not very sensitive to E, the impact properties
were very dependent on E. The threshold energy needed to cause damage depends on the amount of energy that
can be stored elastically and therefore on a'E. Thus usingglass fibres with lower E or using450 layers
on the outside of a laminate to lower the flexural modulus caused a marked increase in threshold energies.
The fracture energy G depends on K'/E so the extent of the damage for a given absorbed incident energy will
be less for a smallerCvalue of E. GRP and aramid fibre reinforced laminates have better impact properties
than CFRP although thdy may not necessarily be tougher. Thus surface layers of GRP or aramid fibres should
produce greater increases in impact resistance than in laminate fracture toughness.
"For laminates 2
with similar values of E, Gcso that the tougher laminates had superior impact
K
z 0
properties, giving greater values of Charpy impact energy, greater threshold energies for damage and
greater residual str-ngths. This was true for laminates 4 to 6, where resin properties were important, and
for laminates 7 to 13, where the fibre properties were more important.
I;
0.
'I 7779
9.5
CONCLUSIONS
Fracture mechanics gave reasonably good predictions of the residual strtngths of CM1P laminates, with
3
fracture toughness between 20 and 5Q MNm_ 2, for notch lengths from 2 mm to over 40 mm, if allowance was
made for the size of the damage zone at the tip of the notch.
The fracture toughness of CFRP and GRP laminates depended on the details of the fracture processes
and hence on the fibre, resin and interface properties, particularly those promoting shear cracking, but
it was relatively insensitive to laminate modulus.
The impact properties were significantly dependent on laminate modulus, the use of GRF or surface
layers of low modulus 450 layers increasing threshold energies for damage, decreasing the extent of damage
and hence giving greater residual strengths.
For laminates with similar values of modulus, the toughness increased wl.th increasing impact
resistance and therefore provided a good indication of :mpact performance.
REPERENCES
G Dorey Fracture behaviour and residual strength of carbon fibre composites
subjected -o inpact loads AGARD Conference Proceedings No 163, Failure
modes of composite materials with organic matrices and their consequence
on design, paper 8 (1975).
"ftI
I I-
9.6
2 T j TABLE
SUMMAPY OF THE COMPOSITE LAMINATES_
RE RED To
Fibres 3 and 4 were carbon fibre with modulus between those of types 2 and 3.
I;
.4
r-- -
7
9.7
TABLE 2
SCharpy
Laminate
-Flexural
Fibre/Resin strength
TTensi le
strength
I
0-1/0ro
Fracture imhacpt
toughness impact
2 2
GN/mr GN/rm MNMm-3/2 kJ/m2
TABLE 3
STRENGTHS AND TOUGHNESS OF (0, 90, +45) LAMINATES
0.8
A Net section strength
i60
Laminate
A A
50-
A A
6 x x
Fracture -x _ A x x 2
toughness 30 x
p"Z MNm 22
20
4_
I' | 1:0N I
0 10 20 30 4.0 50
Notch length mm
I; -
9.9
150 60
Tough
0 fibrous
50
4b/ Fine
inter-
Cha0py too 5a laminar
Charpy
impact
j 0shear cracks Residual I0
Senergy Sb Coarse strength 3
ki 2 xb IL M 2 4a
kJ/X 5b
so cracks 2 a 5
jBrittle o-1\ 6
x Sb
.L I I" ________1__I __ I
0 10 20 30 ,03 0 1 2 3
Fig 3 The Charpy impact energies of some CFRP Fig 4 The effect of drop-weight impact on the
laminates of various fracture topghnes3es residual interlaminar shear strength of
(lay-ups (a) [ (+45)204],, some CFRF laminates.
(b) [0 2 , 2 45
,'-452 0238).
1.2 1.2
1.0 %-" \
\ \
1.0
' \\
,R e s id u a l
0.8 -
. !
\ t
. Re s id u a l
0.8
b ,"
flexural flexural, 5b\
ft~rl 0.60.
stvength
2
strength
2
0.6 _______"__ ""__
GN/mr I N/rn Sa A
0.2 0.2 J
0 1 2 3 4 0 2 4, 6 a 10 1
Fig 5 The effect of drop-weight impact on the Fig 6 The effect of impact by a 6 mm diameter
residual flexural strength of some steel ball from an airgun on the residual t
CFRP laminates. flexural strength of some 0FRP laminates.
.",1
;1 iN
9-10
100
0 E-glass
013
90
Charpy
impact 08 @7
energy
kJ/m 2 Oil @10
80
LF 09
012
70
30 40 so
12 1.2
0 8 - " \0.8 . .
0' 1 2 3 4 50 1 2 3 4
09.1
r p T0.8
ensile stress Zx 20%1
0 0% UTS
0 35%/.
A 500/a
0,6
Residual
tensile
strength 0.,4
2
GN/M
0.2
L________________________.____ ._
C 2 4 6 8
Incident energy J
Fig '0 The effect of bailgun impact on the residual tensile strength of
a CFRP laminate (laminate 14) that carried a static tensile stress
Ouring impact.
Incident energy * 3J
04 0O8.
SResidual j
j tensile
02
[
strength
GN/m
2 /
0 10 20 30 40 50
'1
iJLJ
.". .....
EROSION ET IMPACTS SUR LES PALES D'HELICOPTZRES EN COMPOSITES
par M. TORRES
D6partement Recherches
SNIAS Division Hglicop:-ares
13722 - MARIGNANE
RESUME
Les rotors d'h~licoptares sont soumis a tin certain nombre d'agressions, li~es a l'environ.,ement dams
lequel ils doivent 6voluer. L'utilisation de matfiriaux nouveaux tels que lea composites verre/6poxy et
carbone/6poxy impose au constructeur des programmes d'essais de qualification sp~cifiques a chacune de
ces agressions.
Cet expos6 pr~sente les modes de digUadati >..du t;.pe 6rosion (a la pluie et au sable) et impacts
(civils et militaires) stirles Waes *'.tiw 'omos.tes.
Los essais r~alisgs a la tour, d'grc~.ion ont ?ermis de sAlectionner les mat~riaux de protection les
IIj plus efficaces.
L'exp6rience en utilisatio'n a~nsi que los essais de tirs et d'impacts r6alis6s stir Dales ont montr6.
tin excellent comportement en fatigue apr~s les chocs ou impat, da en grande partie au caractaro fail
safe des mat6riaux composites. Ce bon comportement, associ6 a des poss'lbiliths de reparation tr~s iten-
ii dues, est tin factour important contribuant a la snpfiriorit6 des pales en compo3ites stir les pales
m~toliques.
Los premiares pales d'h6licopt~res en mat~riaux composites produites en s~rie furent cellos du SA 341
Gazelle, d~s 1971. La fibre de verre fut choisie sous forme do rubans pour constituer le longoron, et
sous forme do tissu pour los revitements et renforts. Pu!4 apportant tinallagement notable, 1e carbone Al
haut module fut introduit stir les rov~toments des poles du SA 360 Dauphin en 1974 et du SA 330 Puma en
1975.
L'utilisation do ces mat~riaux permit d'6tendre la dur~e do vie des pales jusqu'A1 tne valour infinie
pratique, et d'accroltre los performances des h6licoptares grace aux possibilit~s do rfia2isation de
formes 6volutives at da profils performants.
I
Les rotors d~h~licopt~res sont soumis a un certain nombre d'agressions, li~es A l'orivironnement dans
jl eq uel i l s do ive nt 6 o u r :-
- Des conditions climatiques tras variables :domaine do temperature s' tendant doe.4 4 7 0 er
70 er
hygrom6trique allant jusqu'a 100 %, conditions givrantes pouvant entratnor des accumulations auibord
d'attaque, possibilit~s do foudroiemont, pluio et grile ongendrant tine 6rosioi. et des impacts dans la
zone du bord d'attaque.
- Des impacts, chocs ou 6rosions, li~s A ltutilisation :chocs en mantutention, collisions, impacts divers
(caillioux, arbres, oisoaux, etc... ), 6rosion auisable, impacts militaires.
L'utilisation do mat~riaux nouveatix tels quo los composites verre/6poxy et carbone/6poxy impose ati rons-
tructeur des programrs dlessais do qualification sp6cifiques a chacune do ces agrossions.
impacts (civils et militairos), lloxp~rience et los essais r~alis6s dans ces domaines, et lea choix
technologiques effectufis pour los protections et les reparations.
.1
1. EROSION
L.es vii esses on extr~mit6 do pales sont 6lov~es :200 Al220 m/s On stationnaire o1280 m/s en
croisi~re rapiae. Les effots do l'6rosion sont donc importants ot los r6perc-ussions air 1e fonction-
nement do l'appareil peuvont itre extrimement n.Fastes
- Diminution des performances par porte des caract6ristiques sarodynamiques des profils,
- Augmentation des contraintes, des efforto do commsands,
- Visillissemont acc~lfr6 des r6sines par l'eau et los rayons ultra violets, iont Ia pfinftration
so trouvo iacilit~e par los piqflres due A 1'6rosion.
L16rosion sur les pales dthfilicopt~res i)eut itre caus~e par dewc facteurs 10lsable et Ia pluie.
Pour 6tudier lea m~canismes do 1'6rosion at 6valuer 105 diff~rentes protections, tin moyen d'essai
particu1'ter a 6tfi mis em place
10-2
D'un diamatre de 4,5 in et drune hati-eur de 4,2 is, elle enferise un rotor bipale de 219 mnde
diamatre.
La vitesse en bout de pale est de 225 m/s, et l'incidence peut itre r~gl~e entre -53et +60.
tinpremier dispositif constitu6 d'une moto-pompe et d'une buse a grand diamatre sitnule une pluie
avoc un d~bit die4 litres/minute * correspondant A un taux de precipitation de 20 rmn/haure. Le
diametre des got'ttes rotor arrete eat de 1,9 a 2 mm.
tin deuxi~me dispositif utilisant un pistolet I d6pression d'air projette du sable de bas en haut,
avec un debit de 180 grammes/minute. Les particules ant un diam~tre cospris entre 0,04 et 0,17 mm.
Les essais pr~liminaires ant montrS que les incidences correspondant au maximum d'6rosion 6taient
de 40 pour les essais au sable et 00 pour les e~sais a la pluie.
La corr~lation entre les essais A la tour et les processus naturels dt~rosion chez les clients est
tras difficile a 6tablir. Ce mayen d'essai permet toutefois d'effectuer des comparaisons enltxe !es
difffrents types de protections.
Quelques r~sultats dlessais stir h~licoptare en environnement naturel naus ont perinis de fixer tine
di.k d'essai a la tour, devant couvrir en mayenne 2000 heures de fanctionnement chez l'utilisateur
A cause des diff6rences de masse et de constitution entre lea gouttes d'eau et !ies grains de
sable, on constate des processus d'6rosion tras diff~rents:
-Les partictiles do sable susceptibles d'9tre sotilev6es par le rotor ant tine faible masse :mains
de 15 milligraines7, -cequi correspond a des diawrtres inf~rieurs a 100 microns.
Leur trajectoire auiniveau du bard d'attaque suit donc partiellement les lignes de courant, et
la zone 6rod~e petit s'4itendre vers le bard de fuite.
Les essais a la tour ant montrg que le long d'un profil, l'intensit6 de 116rosion 6tait fonctian
de l'angle d'incidence, le maximum se situant dans deux zones aitu6es d, part et d'atit' de la
zone frontale (figures 3, 5 et 6).
Par ailleurs, on a pu canstater sur les rotors en utilisation chez les clients qutil existe tine
zone d'6rosion dlnest mod~r~e, de farme triangulaire a l'intrados, en extrimit6, suivarnt
~ une r~partition illustr6e sur la figure 7.
Le m~canisme de 116rosion par le sable eat tel que, pour chaque point de la surface 6rodge, le
taux crerosion est constant du d6btit a la fin de Ilessai, ce taux variant comine le cube de la
vitesse locale.
-A la pltiie, les d~t6riorations localis6es sur la zarq frontale du bard d'attaque sont dues au
martelage des gauttes d'eati ; elles sant constitti6es par des piqr~res otides micro-cratares
rapproch~s (figures 4, 5 et 6).
Coinpte tenti de la grasseur ties gouttes d'eau, lea effets m~caniqties oti a~rodynamiques de ces
d~gradations ioront d'autant plus iinportants qtie lea dimensions de la pale seront faibles.
I ~
Denombreux mat~riaux ont 6tg compar~s au travera d'essais a la tour d'6rosion ; le tableau en
arnnxe pr~sente tine s~lection de risultats d'essais.
A l'analyse de ce tableau, on pout distinguer deux grandes cat~gorios de mat6riaux :les
Les mfitalligues se caract~x'isent par tine bonne tenue a la pluie et une tenue au sable xn~diocre.
Avec ce type de niat~riaux, une 6paisseur minimale doit 6tre respect~e pour 6viter lea risques de
"Imartelage" par.,
lesgouttes d.'eau.
Mais lea difficult~s de formage do ce mat~riau sont tellos qu'il no peut itre Ltilis6
Cotte solution, satisfaisante dana des conditions do fonctionnement normales, ne pout itro
t retenue pour des pales devant voler en conditions givrantes
sont des zones pr~fifrentielles d'accumulation du givre.
en effot, los borduros des cavaliers
Dana ce cas, la coiffo do protection no doit former qu'une pikce, ce que pormetTent do r~aliser
1e titane oti 1e nickel.
Parmi les alliages do titane, le. TA6V pr~sente do bonnes performances A la pluie, mais 1e taux
d'6rosion au sable est prohibitif.
:4 Le Nickel, obtonu par 6lectroformage, so pr~sente counae 1e meillour do tous los mat6riaux
m~talliques ; sa tenue a la pluie est excellentc, et son taux d'6rosion au sable eat do l'ordro
do 0,15 mm/heure a la tour d'grosion.
Un d6p~t 6lectrolytique do nickol sur une criiffe en inox ou en titano pr~sente lea caract6ris-
tiques du nickel jusquIA d6gradat'on do celui-ci, puis celles do la coiffe support.
~ Les thermoplastiques polyur6thanes so caract~risent par une tonue au. sable acceptable et uno
mauvaiso tenue A Ia pluie.
Do noinbreuses oxp6rimentations effoctu~es sur ces produits ont fait apparaltro une dispersion
importante dans lea r~sultata, due on particulior
- au mode de fabrication
- ala duret6
S- a l'6paiassur
- a l'6tat do vieillissoment du revitotaent
- aux contraintos internes r~aultant do l'ernploi de .rove'tements plans pour l'habillago
do formes convexes.
4LCes mat~riaux zont utilisfis aujourd'hui sous douxfom
-bandes do 0,65 mmad'6paisseur, pour des zones a faible vitease (jusquIA 0,8 R). ou pour los
protections d'intrados en extr6mit6.
-bandes autocollantes do 0,4 asmd'6paisseur, coll6os aur los coiffos m6talliques ou I lointrados
en extr~mit4 lorsque l'6rosion au sable eat a craindro.
10-4
Aucun mat~riau n'offre done do protection efficac~e A la fois contre la pluie et le satle.
Les solutions retenues aujourd'hui r6alisent un compromis ac~ceptable, mais suscoptible d'&tre
encore amfiliorg
Pour les pales non d6aivr~es, le bord d'attaque est prot~gg ddns la zone dlextr~init6 (apras 0,8 R
environ) par une coiffe en acier inoxydable 6pais. De l'attache a 0,8 R, la coiffe de bord d'attaque
peut 6tre r~alis~e en inox plus mince ou en polyur~thane.
La zone triangulair- a l'intrados en extr~mit6, expos6e a 1'4rosion au sable, eat recouverte de
polyur~thane. En cas d'utilisation intensive en zone d6sertique, l'applicatlon de bandes auto-
adh6sives de polyurethane, renouvelges p6riodiquem.-nt, permet de prot6ger !eascoiffes en inox.
Pbur les pales d~givr~es, la coiffe de protection, n6cessairement continue, eat r~alis6e en
titane embouti.
1.5. Maintenance
Sur le.s coiffes m~talliques, les rayures, chocs, criquos, ou. d~collements partiels sent admis.
Par contre, le remplacement de la coiffe eat irnpfratif en cas de perforation due a 1l'6rosion ou A1
des impacts, ou en cas de d~collements importants.
nI 2. CHOCS-IMPACTS
Les pales d'h~licopt~res peuvent subir au cours de leur vie diff6rents types de dormnages
- chocs, rayures, au cours des op6rations de manutention,
- chocs et impacts en exploitation :heurts d'arbres, impacts de cailloux ou dtoiseaux,
collisions,
-impacts militaires.
Les matgriaux composites ont permis d'am~liorer consid6rablement le comporternent des pales d'h~li-
I copt~res dans ce domaine :grace A leur caract~re fail safe, las degradations restent limitfies et se
propagent lentement ;par ailleurs, grice I leur faible densit6, les 6paisseurs de zmat~riaux mises en
jeu dans la zone du bard diattaque sent souvent plus 6lev6es que celles des mat6riaux m~talliques, ce
qui a pour effet d'accrottre la tenue A la compressilon, donc la tanue aux chocs frontaux par exemple.
Ces propri6t~s peuvent ktre illustr6es par un certain nombre d'exemples
2.1. Chocs et impacts en exploitation
Pour verifier le comportement des pales principales dans cette situation, des essais specifi-
ques ant 6t6 r~alis~s. Ces essais, r~alis~s sur un rotor de SA 341 Gazelle, Dint consist6 a
introduire des branches d'un diarnatre croissant dans le rotor.
-Des branches de bais vert de diam~tre 220 mm puis de bais sec de 200 mmant 6t6 introduites
(figure 8) :on observe alors des d6gradations sur le saumon d'extr6mitg et des d~collements
de revgtement au bard de fuite. Ces d~gradations n'ont pas d'influence sensible sur lea
performances, clest-A1-dire que la sustentatian et le contr~le de l'appareil restent assur~s
apras Ilimpact ; bien entendu, le niveau vibratoire se degrade, mais ce type d~impact aurvient
n~cessairement au voisinage d'i sol et peut done atre suivi d'un atterrissage immhdiat.
10.5
tin client, lors d'un atterrissage dienuit en brousse avec le SA 330 Puma, a heurt6 avec
ses pales principales en composites (verre et carbone/6poxy) des arbres dont le diamatre 6tait
d'environ 200 a 400 mma.Malgr6 cet incident, l'appareil a poursuivi sa mission pendant 45 minutes.
Au retour a la base, on observait sup trois pales des d6gradations traversant le rev~tement et le
remplissage nid d'abeille diel'intrados a l'extrados, stirdes surfaces de 300 x 300 mm environ
stir la qtiatri~me pale, on notait des d6gradations plus importantes en surface avec rupture de
l'aratier diebord diefuite.
Apr~s ilexpertise, seule la quatri~me pale a 6t6 rebuttie, et lea trois autres ont Pu 6tre
r6pardes (,Figure 9).
Dana des circonstances plus malheuretises, en Novembre 1975, tinn6licoptko SM 341 Gazelle entra
en collision avec tin avion du type BONANZA. Malgr6 dea d6gradations impor1.ertea stir le rotor,
l'h6licopt~re put attevrir et l'.5quipage sortit indemne diecat accident (Itig.ure 10).
r Lea essais de fatigue sont effectti6s aur des banca (figure 11) capables dieaimuler, avec la
force certrifuge, des charges diebattement et de train6e die3 a itfois lea charges maximales die
I ~vol.Ces installations, propres a chaque type de pale, permettent dietester s6par6ment la zone
d'attache otila zone diepartie courante.
Sur SA 341 Gazelle, des tirs ont 6t6 offectugs aux calibres 20 mm explosif, 12,7 mm normal,
et 7,62 mm NATO non perforant.
Lea tirs ati calibre 12,7 mm ont donn6 les r6stiltats suivanta le premier, perpendictilaire aui
plan diela pale a 90 rumdu bord d'attaque, n'entraine pas de dommages importants ;le deuxia~ie,
dens le plan diela pale et face atibord d'attaqtie, a d6grad6 tinepartie du lon;eron 3:Lnsi que le
bord diefuite. tinessai statique a montr6 que la tenue de la pale 6tait encore sup6rieure aux
efforts centrifuges.
Les tips ati calibre 7,62 mm (figures 12 et 13) ont 6t6 effectu6s stir deux trongons diepales
destin6s a 6tre test~s ensuite en fatigue :le tir stir le trongon de partie courante 6tait perpen-
diculaire atiplan de la pale, en plein longeron, dana la zone la plus charg6e en contraintes ; le
tir sur le trongon d'attache 6tait dirig6 perpendiculairement a ltaxe diela pale, a proximit6 des
douilles d'attache.
Lea essais de fatigue, effectu6s aux charges maximales dievol, on,.;t6 conduits jusqti'a l'6qui-
valent de 3 heures dievol pour le trongon dtattache, et 10 heures pour celtii de partie courante,
sans 6voltition des d6gradations autour des impacts.
Ces essais ont donc permsi de dfimontrer qu'tin appareil ayant tine oti plusietirs pales endommag6es,
mime dana lea zones critiques, petit regagner sa base.
Sur SA 330 Puma, 1:: tirs au calibre 12,7 mmperforant ont k:6 eff::t:6s stir trois trongons de
le pemir
tr tavesai pepen uairt
emtl trav~nersai t i orneanrir d
longeron ~ ; l'essai
dattqu fatigue, conduit
die en battement rWA1fq
pra6t lent
taine 2
voli.ln
ti arrrt6 d 6hersdevl
le toiske
ir tai diig6sur lea douilles d'attache ; l'essai diefatigue, conduit en
battmenattr~n~eave
foce entifue, 6t arit6a 116quivalent de 6 heures dievol.
6
Ads toisessisaucne6volution notable des endommagements n'a fit6constat a.
Iarrt
-4a
10-6
La fin d'une mission est donc garantie dana tous les cas de vol, au w'me risque que celui
autoris6 pour la certification de pales principales, qui eat d.a10 -6.
3. MAINTENANCE - REPARATIONS
Outre .linspection quotidienne ou a chaque vol 9 une d6pose des pales eat prescrite routes lea
100 ou 300 heures suivant les appareils. Au cours de ces d~poses, lea pales subisaent deux contr~les:
contr6le visuel de d6tail et rontr&.. acoustique ("taping,:), permettant de met tre en 6vidence lea
d~collementsaet les d~laminages.
Un certain nom~bre de d~fauts ou d6gradationa peuvent 9tre admis sans r6paration :rayures ou
chocs sur le revetesent en composite pr6sentant des endomriagements de xaiblea dimensions (0,1 a
0,3 immde profondeur, 20 mmnde longuour), criques, rayures, ou impacts sur lea coiffes rn~tall3.ques,
d~colle.'nenta du revitement limit6s en nombre et en surface, etc...
Lea d6gradations plus importantes n~cessitent une r~paratiun. Les limites des d6fauvs admissibles
et lea r~parations sont parfaitement d~finis et sun juatifi~a par le conatructeur grace a des essais
en fatigue sur des trongons de pales sp6ciaux.
Deux niveawc sont alors distingu~a ; lea r6parationa ex~cutables chez le client, et celles
n~cesaitant un retour en usine ou chez un r~parateur agr66.
I
Lea r~parations importantes doivent atre effectu6es en usine ou, pour certaiaes, chez un
r~parateur agr66 remplacement des coiffes m~talliques, reconstitution du revatement par empiace-
ment ou 6clissage, reraplacement d'une zone de remplissage (mousse ou nid d'abeille), r~paration
de la languette de bord de fuite par addition de cavaliers en tissu de verre/6poxy, reconstitution
d'une partie du longeroa , etc...
Lea possibilit6s de reparation des pales en composites sont tr~s grandes, et souvent seules
des considgrations de prix de revient limitent l'imporvance des r6parations effectu~esaen usine.
4. CONCLUSION
Les essais r~alis~s A la tour d16rosion ont 'dsaen Svidence les bonnes performances a Ia pluie
de l'acier inoxydable, du nickel, at du titane, ainsi que la sup6riorit6 du polyurethane pour
1'6rosion au sable.
Des incidents en exploitation et des essais sp~cifiques ont rDntr6 l'excellent comportement
au choc des pales en composites, a l'occasion d'impacts d'arbres d'au momns 200 mmde diamktre.
Dans Ie domaine militaire, lea essais de fatigue r~aliske apras des tirs au caiibre 12,7 mm
dana lea zones critiques de la pale ont montr6 que l'appareil pouvait dans tous lea cas rejoindre
Ce bon cornportement aux chocs et aux impacts da au caractare fail-safe des ma 6riaux composites,
associ6 A des possibilit~s de reparation tras 6tendues, eat un facteur important contribuant a
lasup6riorit6 des pales en composites sur lea pales m6talliques.
10-7
IP
.PLUIE
,-, : -.
-SABLE
POSITION EN COIIDE
0
EXTRADOS 100/0 50/0 0 0 /o 5 /o 100/o NTWLNRAQOS
tPLUIE MNUI
____ SABLE
i-+ so
i= 00
INTRADOS
1.
J, .~
Figure 8 Rupture d'une perche de bois Figure 9 Pale PUMA d6t6rior6e, avant
Sde 200i'nn x i 200 nn par des pales Gazelle. rparation.
;
!0-0
I-k
Figure 10 SA 341 Gazelle apr.s une Figure 11 Essai de fatigue sur pale
collision avec un avion 16ger Puma
4 , "
Wc-1
ANNEXE
ESSAIS D'EROSION
POLYURETHANES 1 a5 0,05
II
.4
2 '
J,
lI -I 2
by
SUMMARY
The effects of compression load intensity were such that lightly loaded graphite
structures (such as aircraft secondary structure) were insensitive to impact damage. In
more heavily loaded structures, (such as wing panels), however, appreciable reductions
in compressive strength occurred. The implications of the tests for structural de3ign
are discussed by comparing panel masses for designs where altimate strains have been
reduced due to impact considerations with the masses of designs with higher ultimate
strains. Finally, preliminary test data is presented to show the possibility of improve-
ments in damage tolerance achievable by using an alternate matrix material.
INTRODUCTION
TEST SPECIMENS
Three types of test specimens were used in the investigadon: sandwich beams,
laminated plates, and stiffened panels. In Figure l(a) a typical graphite-epoxy
sandwich beam is shown. The back face of the specimen is a .32 cm thick steel plate. -j
The sandwich beam specimens were loaded in a four-point beam bending fixture to obtain a
I A
11-2
uniform compressive or tensile loaded test area 7.6 by 7.6 cm in the middle of the
specimen.
A typical laminated plate test specimen is shown In Figure l(b). The plate is
simply supported on its sides by two adjustable edge supports and clamped on the top
and bottom ends where load is applied.
A typical stiffened panel specimen is shown in Figure 1(c). Parels were either
hat.-stiffened or blade-stiffened plates that were designed with minimum mass to carry
specified compression loads over a specified panel length. The ends of the stiffened
panel were potted in a filled epoxy rzsin while the sides were unsupported-
The epoxy materials used in specimen fabrication were commercially availible 450K
cure thermosetting systems. The graphite fiber was a high clrength continuous filament
type and was used in both unidirectional tape and bidirccuional balanced weave fabric
forms. A test matrix for the test specimens is pe.esented in Table I indicating
principal test objectives. Further details can be found in Refs. 3 to 8.
TEST TECHNIQUES
Impact Apparatus
Aluminum spheres, 1.27 cm in diameter with a mass of 3 grams, were used as the
impact projectiles. An air gun was used to propel each projectile. As illustrated in
Figure 2(a), compressed air is used to pressurize a reservoir, until a thin Mylar
diaphragm at one end of the reservoir ruptures. The air in the reservoir is then
metered through an orifice plate and subsequently drives the projectile down the
barrel. An electronic velocity detector at the end of the barrel is used to measure
projectile velocity. The projectile passes through two light beams which trigger
electronic timing equipment to record the travel time between beams. The impact
conditions were selected to simulate energies from small rock or hailstone damage (such
as that occurring during an aircraft take-off or landing) to large hailstones and
dropped tools.
The test setup for a typical panel is shown in Figure 2(b). In this case, the
panel shown is loaded in compression by the platens of a testing machine. in all tests
the impact was normal to the specimen surface. Specimens were damaged by impact at
different velocities while subjected to static load to determine the combination of
impact and load necessary to cause failure. Specimens which did not fail or which were
damaged without load were statically tested to determine the effect of local damage on
residual strength.
1
stiffened panel specimens, the presence of the sandwich core appeared to confine damage
(prior to failure) to this local area.
The appearance of damage in graphite-epoxy plate specimens is shown in Figure 4(b).
o_ The first visible evidence of impact damage is a crack on the back surface of the
specimen as shown in the lower left photo of Figure 4(b). For the specimen shown that
=mm
T -- ......... ... -. .;4
11-3
waE, impacted at the lower velocity, there is no visual indication of damage in the con-
tact region. At higher impact veloiities, the side opposite the impact location has a
large region with surface (racks (right side of Fig. 4(b)) and portions of the surface
ply have either delaminated or spalled off. T"hneonly evidence of damage in the region
of contact was a shallow circular depression about two-thirds of the projectile
diameter. Similar observations were mtad) on stiffened panel specimens. All plate and
stiffened panel specimens mad a largtir region of visible damage on the specimen surface
opposite the impact location than in the bontact region.
Interior Damage
Damage Sequence
The damage pattertis observed suggest a possible sequence for laminate damage to
occur due scr .. w-velocity impact. As illustrated in the schematic drawing of Figure 6,
local fai~ureo are created within the laminate and are probably due to reflection of a
stress wave from the back surface. A specimen with such damage would have a reduced
local bending stiffnesz and could deform locally out-of-plane as the projectile
decelerates. Such local distortion was observed on several plate specimens. The
superimposed compression load could couple with the local out-of-plane deforma~ion to
cause the damage to propagate or to cause additional delamination and intruply cracking
to develop.
damage. The x-marks on Figure 7(a) indicate the residual strength obtained 1rom the
three beams which did not fail on impact. The data shown in Figure 7(a) is for a graphite-
I epoxy beam made of 181-style cloth with a [0 J layup. Results obtained gor other orien-
tations using graphite tape, Kevlar, and Kevlar-graphite hybrids are similar and csn be
found in Reference 5. Comparable results were observed for beam laminates loaded in
tension, although damage propagation mechanisms are probably different.
In Figure 7(b), typical results for laminated plates are presented. The failulr2
.~* threshold curve shows that plates loaded with compressive strains below about .003 strain
4 are insensitive to impact d.mage. For h!,her load:, serious reductions in strength may
-occur. The data shown in Figure 7(b) is for a 48-ply (+45/0 2 /+45/0 2 /+4 5 /0/ 9 0) 2 s graphite-
-epoxy laminate. These trends were found to b? typical for srveral ply orientations, and
mixtures of taoe and cloth. (See Refs. 6 and 7.)
eachIn Figure 7(c), typical results for graphite-epoxy stiffened panels are shown. For
each of three designs, the applied panel compressive strain at failur'p is compared with
the design stra!i, at buckling denoted by the dashed lines. The lightly loaded panels
(Design A) achieved their design strength in the presence of impacts of 51-54 m/s on
"critical stiffener locations, and were insensitive to low-velocity impact damage. The
heavily loaded dec.sgn (Design B) experienced serious strength reductions when impacted on
stiffeners at velocities ranging from 51 to 03 m/s, and hence were found to be damage
0-1
11-4
In Figure 9(b),curves are presented for graphite-epoxy panels with maximum design
strains of .0060 and .0030. The lower strain value corresponds to a safe strain level
from an impact damage point of view, whereas the larger strain value is a maximum based
on matching the bending and torsional stiffness of an existing aluminum wing. (Matching
stiffnesses reduces qualification problems for the retrofitted composite wing). The
graphite panel results can be compared with the cross-hatched curves representing the
mass requirements for current aluminum commercial aircraft wing panels. For lightly
Nx
loaded panels .77-Mpa.), a mass-saving of nearly 50 percent is predicted for
graphite-epoxy designs versus aluminum designs. For heavily loaded panels, E- - 5 Mpa)
savings are about 40 percent for a maximum strain of .0060, and are reduced to about 15
percent when impact sensitivity is accounted for (cmax 3.0030).
The calculations presented for graphite panels were obtained from a recently
developed structural panel sizi,g computer code (Ref. 9). The calculations were con-
stratned to match extensional and shear stiffness properties of an equivalent aluminum.
wing panel structure, and included minimum gage constraints and effects of an overall bow
irnperf,,tion.
preliminary data shown. Thus, it appears that certain matrix materials may be -/
"tougher" in their resistance to impact. In current research, efforts are directed
S~toward understanding what mechanical and chemical properties are critical in enhancing
the damage tolerance of compression structures.
The traditional Charpy and Izod impact tests have been used by a number of investi-
gators to evaluate the impact response of polymer-matrix composite materials. As
discussed in Reference 2, however, data from impact tests can provide conflicting results
and interpretations. For example, Izod impact tests (Reference !0) conducted on the
same material systems as those (Materials A and B) of Figure 10 1-ndicate that Material A
exhibited a 45% improvement in impact performance over Material B. By contrast, the
low-velocity impact data obtained using ball-type impacts on compression-loaded specimens
reported herein, clearly shows the superior impact resistance of Material B over Material
A. (See Fig. 10)
It is believed that the type of impact test performed in the present investigation
has substantial merit. First, the test method is simple ana requires no sophisticated
equipment to perform. Second it is more realistic in representinv impact effects on
in-service structures where in-plane loads and panel response must be considered. Third,
the test gives direct information on the reduction in material strength and consequently
the data can be used directly in design applications.
CONCL.UDING REMARKS
Typical results nave been presented from an extensive investigation of the effect
of low velocity impact on the compression strength of graphite-epoxy structures. Data were
obtained in this investigation on graphite-epoxy sandwich beams with thin facesheets,
48 ply laminated plates and stiffeneu panels designed for minimum mass to carry specified
compression loads. Test results suggest that lightly loaded panels which may be
designed by stiffness requirements may achieve their design strength in the presence of
impact damage. Heavily loaded, strength-critical panels, however, experienced siguifi-
canc strength reductions due to impact damage that is not visually detectable.
Observation of the cross-section of impact-damaged specimens suggests that damage is
initiated by stress ayes and that local out-of-plane deformation caused by the
proj.ectile may propagate the damage or cause additional interior damage to develop. The
compression strength of impacted test specimens has been shown to be related to the
applied in-plane compression strain when impact occurred.
Current aluminum commercial aircraft wing panels are designed by material strength
and have design ultimate strains we2J above the impact-sensitive strain levels observed
for compositss in this investigation. Studies indicate that stiffened composite
compression panels designed to match extensional and shear stiffness properties of
aluminum wing panels with composite strain allowables limited by impact damage offer a
potential mass savings of at least 15% when compared to existing aluminum designs.
However,
materials savings
can be of up to 40% may be realized
improved. Preliminary studiesif suggest
the damage
that tolerance of graphite-epoxy
impact damage tolerance may
be enhanced through improvements in the epoxy matrix material. Further research,
however, is required to understand the critical chemical and mechanical properties of
the matrix nec,-ssary to achieve the desired level of damage tolerance in composite
compression structures.
REFERENCES
1. Anon: Standard Methods for Notched Bar Impact Testing of Metallic Materials, ASTM
E23-72 (Reapproved 1978). 1979 Annual Book of ASTM Standards Part 10, pp. 237-253.
3. Rhodes, M. D.; Williams, J. G.; ard Starnes, J. H., Jr.: Effect of Low-Velocity
Impact Damage on the Compressive Strength of Graphite-Epoxy Hat-Stiffened Panels.
NASA TND-8411. April 1977.
4. Rhodes, M. D.; Williams, J. G.; and Starnes, J. H., Jr.: Effect of Impact Damage on
the Compression Strength of Filamentary-Composite Hat-Stiffened Panels. Paper
presented at the 23rd SAMPE National Symposium and Exhibition. Anaheim,
California. May 2-4, 1978.
05. Rhodes, M. D.: Impact Tests on Fibrous Componite Sandwich Structures. NASA TM-
78719, 1978.
*i' 6. Starnes, J. H., Jr.; Rhodes, M. D.; and Williams, J. G.: The Effect of Impact
Damage and Circular Holes on the Compressive Strength of a Graphite-Epoxy Laminate.
NASA TM-78796, 1978.
11-6
7. Rhodes, M. D.; Williams, J. G.; and Starnes, J. H., Jr.: Low-Velocity Impact in
Graphite-Fiber Reinforced Epoxy Laminates. Paper presented at the 34th Annual
Cnference Reinforced Plastics/Composite Institute. The Society of the Plastics
Industry, Inc. New Orleans, Louisiana. Jan. 29 - Feb. 2, 1979.
8. Williams, J. G.; Anderson, M. S.; Rhodes, M. D.; Starnes, J. H., Jr.; and Stroud,
W. J.: Recent Developments in the Design, Testing and Impact-)amage Tolerance of
Stiffened Composite Panels. NASA TM 80077. April 1979.
9. Anderson, M. S.; and Stroud, W. J.: A General Panel Sizing Computer Code and Its
Application to Composite Structural Panels. AIAA Journal, Vol. 17, Number 8,
August 1979. pp. 892-897.
10. Riewald, P. G. and Zweben, C.: Kevlar 49 Hybrid Composites For Commercial and
Aerospace Applications. Proceedings 30th Anniversary Conference Reinforced
Plastics Composites Institute, The Society of the Plastics Industry, Inc.,
Feb. 4-7, 1975.
STIFFENED 30.5 - 76.2 x GRAPHITE-EPOXY 60 50- 100 -HAT STIFFENERS VS. BLADE
PANELS 22.1 - 81.3cm BORON-EPOXY STIFFENERS
DESIGN LOAD: -SKIN VS. STIFFENER
.53 -5.8 MN/m IMPACT
"IWDTH &LENGTH EFFECTS
: "DES IGN STRAIN
'4)___ __ _CONDITIONS
4r
4S"',
S to% ,I'
up
11-7
A) Sandwich Beam
Figure 1 - Test Specimens
IN
t (.
I,. C) Blade-Stiffened Panel
-iV Figure I - Concludcd
11-8
ELECTRONIC COUNTER-\ __
DIPRAMORIFICE PLATE
SPACER -
LIGHT EMITTING
DIODES
AIR GUN VELOCITY DETECTOR
050
FIERS0F:0A 900 FIBERSQ
Ia
Figure 3 -Concluded
impat17.6M/S
a a Seedof mpat ata Seedof 136.6 M/S
B) Dmag
4pearace n PateSpecimen
A) Impact Veloc,.ty =62 M/S
11IkV17
I 4C ~,1
TEST I
SPECIMENROJECTILE
1.0 -* FAILURE
0 NOFAILURE
SRESIDUAL STRENGTH
.8 VISIBLE
SURFACE
DAMAGE
.6 .
.4 0 0
NOVISIBLE
SURFACE DAMAGE
.2
K TEST SPECIMEN
38. 1cm BY25. 4 cm
12.5 cm BY 25.4 cm
.01
O 0 FAILED ON IMPACT
B) Laminated Plate
1
.OOB 0 0 DIDNOTiFAILON IMPACT (+4j5/0 2 /+ J5/0p/+415/0/90)~2S
SI .002 -0 I I 00I I
0 25 50 75 100 125
PROJECTILE VELOC I1TY.
ml s
I I I I I J
4 8 12 16 20 24
PROJECTILE ENERGY. JOULES
HEAVILY LOADED
GRAPHITE PANELS (B)
.0 Al A2 A3 BI B2 B3 B4 85 B6 B7 C1 C2 C3 C4
.4i TEST PANEL
u 8
j il
.006 --
.005 -
IMPACT
DAMAGE-SENSITIVE GRAPHITE/EPOXY
.004REGIME ULTIMATE
UPPER WING TRANIN
T BASED
ON
SURFACESTRAIN ISPA IMPACT
F-n
R I I I
.002 A0
.001 '
0 .2 .4 .6 .8 1.0
WING COORDINATE, n
A) Compressive Ultimate Strains for a Commercial Aircraft Wing Structure
Figure 9 - Effect of Strain Limits on Panel Design
50
MAXIMUM DESIGN
STRAIN
NEL
0.003
30 -ALUMINUM WING
jW 0.006
C I I
ECA
MI AIRCRAF I I I I I
".4 -i'8- MPa
B) Effect of Design Strain on Panel Structusal Efficiency
Figure 9 mpConcluded
STRAIN 0 12
STRAIN, %
.006 0
A 0
0
0
V0 0
.002 00 FAILED ON IMPACT 100 MIS
0 0 DID NOT FAIL ON IMPACT
0 4 812 16 20) 24
IMPACT ENERGY,J
Figure 10 -Damage Tolerance Improvement Using Tougher Matrix
12-1
ABSTRACT
INTRODUCTION
No thing is perfect. For design, quality control and maintenance of an aeroplane the question comes up:
which kind and size of deviation is acceptable, which anomaly must be olterd so that the structure can
do the job it was assigned for. To enable that decision, it is necessary to know something aboLt the
behaviour ot the observed anomaly, about its influence on the performance of the structure.
composite material:
(.1. To minimize the inducements of damages. This is mairly a problem of the manufacturer and the user of
a structure.
2. To get informations about the inherent ability of endurance of impact loads without damage as well as
the ability to absorb energy by a fracture process. Since these properties sometimes are incompatible,
a rank must be settled depending on the p, rpose of the structure. E.g. for a skin of an outer wing
section it seemns to be more important to withstand an impact of runway debris without any damage
than to stop a high energy object by means of absorbing fracture energy. lhe opposite might be true
for a fuel tank.
3. To get information about the influence of a specific damage on the performance of the structure. Since
the considered mechanical damages mustly do not initiate the total breakdown of the structure, we have
to find out the damage propagation depending on the service loads. Simplified spoken, this is a fatigue
r: problem.
,~
Minimizing the influence of damage requires consideration of all three items, Of course, the available
capacity in the Institut fUr Bauweisen- und Konstruktionsforschung urged a selection of our activities.
r 11 At the beginning of our involvement in damage tolerance problems we composed a list of possible inducements
of damages in composite materials either by overloading or human error, only as an aid for determination
of our future work. On the other hand our first effort was focussed on the damage process in a real
12-2
structural element, the leading edge of the ALPHA-JET horizonta" stabilizer, impacted by runway debris or
hailstones. After all, we confined us to the investigation of mechanical damages, especially of impact
damages, caused 1y dropped tools, flung runway stones and hailstones.
At present, the effort in our institute is directed towards gaining information about the influence
of these mechanical damages on the lifetime of dynamically loaded thin shells made of composite material.
This paper mainiy deals with the ongoing experimental investigations in our institute.
A) Impacted Beams
A hard object hitting a composite structure with finite velocity, will induce a visible, non visible
or no damage. Earlier efforts were concentrated on the total fracture energy absorption during impact
a similar procedure for metallic materials with small changes only. Obviously the aim of earlier
investigations mostly was to maximize the total fracture energy. Seldom attention was payed to the
elastic, reversible energy absorbed withovt damage. Both the !Lod and Charpy impact tests seem not to be
an adequate method to gain beneficial inft-mation about the damage in thin plates impacted by dropped tools,
stones or hailstones. Stress state, impactor mass and velocity are usually extremely different. As an
example Fig. I shows the influences of the length-to-thickness ratio and the static versus dynamic loading
in three-point-bended beams reinforced with two different fibre types orientated parallel to the main
axis. The bending stress and the interlaminar shear stress at failure as well as the fictitious modulus
and the elastically endured energy depend strongly on these parameters. The bending stresses shown in the
graphs are corrected for large deflections, resp. high length-to-thickness ratios (Fig. 2a). It is easy
to explain the shape of the apparent initial modulus of elasticity taking into account the short beam
shear deformation (Fig. 2b). But that is not true with respect to the strengths. A failure hypothesis
which describes the observed stresses at breakdown of the specimen is not known,
The merit of this test Is to give a rough idea of what happens if this material is subjected to tension
or compression stresses AND shear stresses AND surface pressure simultaneously for a short time.
I The most important experiences delivered from Charpy impact tests which are perhaps transferable to
impact loaded plates are the following:
Glass fibres endure more elastic energy and absorb more total fracture energy than carbon fibres,
aramid fibres and boron filaments under most load combinations. Unfortunately, one of the most important
requirements, stiffness, is not attractive with glass fibres comparel to other fibres. The available
compromise reached e.g. by mixing glass and carbon fibres is not always a qood compromise since in many
! cases the negative properties (e.g. bending compliances) are more dominant in the mixture. Fair compromises
are possible about fracture energies and elastically endured energies as presented in Fig. 3. Those
graphs show Charpy impact test data of unnotched specimens composed of different fibres with several
stacking sequences.
Charpy tests point out the outstanding role of shear strength for the damage mechanism at impact
loads. Even for a length-to-thickness ratio of 40 the beams failed also due to interlaminar shear, not
only when statically but also when dynamically loaded even at rather low nominal shear stresses. This is
reported by SHLENSKII 110) too. BADER et.al. (2) have shown that the relation between endured elastic
energy and total fracture energy depends strongly on the intralaminar shear strength resp. the interfaclal
bonding.
A;
There are many other features.observed on Charpy impact specimens, e.g. the loading rate dependence
of interlaminar shear strength pertainiig to multidirectional lay-up reported by SAYERS & HARRIS (8) as
well as by MALLICK & BROUTMAN (7). In our opinion, these features are related too strongly to the special
12-3
test conditions, this makes it often very difficult to extend those beam results to impact loaded plates
and shells.
The significant characteristic of a plate compared to a beam is its two dimensional extension, that
means an impact load is parried by primary stresses in three dimensions. This seems to he more complex than
in a beam. But in fact it should be easier to understand the stress state in a plate except at the edges
than in a bean (a degenerated plate) since the real three-dimensional stress state in a beam is mainly
governed by the secondary edge stresses. Nevertheless a theoretical discourse on impact loaded thin
orthotropic plates including surface pressure does not exist. Experimental results presented in the
literature and achieved by own work - see the scheme of our test equipment Fig. 4 - may allow the following
statements, which are of general but of course not absolute validity.
A very important parameter is the lamina or fibre orientation. Unidirectional (UD) lay-up is
unacceptable in most cases. One or few very long cracks throughout the laminate would occur even at very
small loads. Already shorter cracks form crosswise in a simple 0/90 0 lay-up or a large zone of sequential
delamination will appear as reported e.g. by SIERAKOWSKI (11i. The least extendion of cracks appears
generally in a multidirectional lay-up excepting the outmost UD layer. Fig. 5 illustrates some of the
statements mentioned above. These long cracks in the outmost UD layer can be impeded by a thin cover of
glass or aramid woven fabric. Fig. 6 demonstrates the difference in response of a covered and a non
covered laminate to impact by a dropped tool. BEAUMONT, RIEWALD, ZWEBEN (3) and DOREY (4) reported on
similar observations.
Concerning the stacking sequence in a 0/90o lay-up SIERAKOWSKI et. al. (11) observed a lower first
failure energy '.nda lower absorbed fracture energy in gfrp composites with alternating stacking sequence
containing many sheet boundaries, than in a three-laminae symmetric lay-up. A rough calculation of the
stresses cou'd not explain this behaviour. Additionally they reported on a sequential delamination
mechanism in 0/9g0 gfrp, which we observed in cfrp laminates at low impact velocities,
0/900 too.
The most influential impact parameter is the projectile velocity. The damages at different
velocities are shown in Fig. 7 for an E-glass laminate with good fibre-to-matrix bonding, With less
interface bonding and low velocities the failure mode shifts from fibre crack to interlaminar and
4iI intralaminar deboncing, nevertheless the dependence of damaged area size on impact velocity is similar.
Higher velocities result in larger damage zones up to a velocity where the damage begins to be
characterizd by a punch-out mechanism. In a limited range the damaged area is linearly related to the
kinetic energy as reported e.g. by SIERAKOWSKI (11) and DOREY (4) at their test conditions. Evaluation
of Fig. 7 (see the scetch in Fig. 8) confirms this experience. Beside the material parameter this velocity
dependence is affected by the support boundary conditions which influence the internal dynamic stress
state of the loaded plate (Fig. 8, below). In a sufficiently large plate the deformation mechanism depends
on the impact velocity as shown principally in Fig. 8. At low impactor speeds the plate behaves like
statically loaded. The energy is stored, mainly near the load introduction at the impact locus and the
supports, that means in a relatively large material volume. At very high speeds the punching affects
an area only slightly larger than the cross section of the projectile. At moderate velocities a mixture
of these two mechanismsAs evident. At the moment of maximum plate deflection the deformation exists only
in a limited area, the extension of which depends on the impactor velocity, mass and stiffness relation
.- 4 between impactor and target and the shell curature and/or support conditions.
Concerning this punch-out mechanism, we could drasticly prove the outstanding advantage of glass
(5 fibre fabric over carbon fibre with a real application. A variant of the leading edge of the ALPHA-JET
horizontal stabilizer was made of pure SC 130/Code 69 material in a 01900 lay-up, 3 mi thick, covered
with a 0,2 mm thick aluminum foil. It was impacted with a glass sphere (13g, simulating idealized runway
{ .x
debris) at v = 55 m/s (approx. the landing speed). The response (see Fig. 9) was a penetration, One of
the investigated variants, 2,5 mm thick, consists of about 70 % glass cloth style 181 / Code 69 and
30 % sc 130 / Code 69 in the midplane in 0/900 orientation. At the rear side of the impact locus we
12-4
observe delamination and deformation. The front side shows no visible marks except some whitening which
would he invisible under a protective painting [1, 12).
On the whole, the possibilities of designing thin shells for good impact resistance are not very
numerous since the most jnfluential parameters are dictated mostly by stiffness, strength, aerodynamics
or fabrication requirements. That means, since it is often difficult to optimize a laminate composition
wi'.h regard to impact resistance, the best we can do seems to gain knowledge about the effect of damages
which can not be avoided in commonly used laminate types, and to give recommendations for modest
alterations of those laminates to improve damage tolerance without weakening of other specifications
significantly.
EFFECT OF DEFECT
In our institute we began to investigate the consequences of an impact damage in composite material
especially with respect to the lifetime of the damaged structure. Principally, we try to find the
influence of mechanical damage caused by a dropped tool to the resistance of glass and carbon fibre
reinforced epoxy on harmonic cyclic loads.
The first preliminary test results are presented below. We started with glass fibre fabric
reinforcement. Three kinds of damages were rendered:
a) ball impression onto a rigidly supported laminate with a squeeze force of 5 kN and a ball diameter
of 11 mm which results in a visibly damaged area of aboug 30 mm2 ,
b) a threesquare indentation into a rigidly supported laminate with a punch force of 5 kN and 6 mm
punch diameter which resulted in a visible damage of about 30 mm',
c) a ball impact onto a plate cldmped in 130 mm diameter .with a glass sphere, m = 30 g at v = 45 m/s
which resulted in a visibly damaged area of about 50 mm2 or 500 mm if scattered small interface
cracks are taken into account, too.
The pictures in Fig. 10 (middle) illustrate some of these damages in gfrp. The fatigue curve of
damaged and undamaged specimens loaded dt R = 0.01 (zero-tension) is presented in rig. 11. The few
available test data indicate only a small variation in the behaviour of damaged and undamaged samples.
The difference between the two matrix systems results primarily from different fibre volume caused
by different fabrication methods (prepreg versus wet lay-up). Glass fibre fabric seems not to be sensitive
to those damages w had introduced and tested at harmonic cyclic tension loads. It may be noticeable
that the type of failure mode changes from a shear type failure at high loads and a small number of
cycles to a tension type failure at low loads and a large number of survived cycles. At high to moderate
load levels the defects behave somehow like a drilled l'ole inspite of remarkable delamination propagation
in the course of the loading history (Fig. 12, above). The final break down usually occured all at once
without any noticed warning. Two intact specimens were tested for residual ouasi-static tension strength
after about e onillion cycles at low cyclic tension stresses. There we observed for the first time a
remarkable amount of the residual strength we registered very often later on with cfrp.
Further iavestigations in our institute were concentrated upon carbon fibre reinforced composites.
4! Up to now we explored T300 in Cnde 69 from Fothergill & Harvey received as UD prepreg sheets. A small test
* {series was run using balanced T300 cloth in L02/SL from Ritag. These samples consists of 5 layers and
were 1,35 mm thick at a fibre content of about 45 vol.%. The corresponding lay-up composed of
* unidirectional prepreg sheets (0/90)3s was 1,7 mm thick at 54 vol.% fibre content. The most extensively
used stacking sequence (02/+/- 45/02 /+ /-45/90)s wes 2,5 mm thick at 55 vul.% fibre content.
I ________ _______
12-5
The damaged and for comparison undamaged specimens were fatigue loaded or tested for residual
strength after dynamic iuading acco-ding to Tab. 1. The quasi-static strength is ircluded as a special
case of fatigue strength.
The results of fatigue strength and residual strength tests with indentation-damaged and intact cfrp
samples re presented in Fig. 13. They are not proved very well ind consolidated in detail since the
available number of specimens and machine time was limited an! we preferred to gain a broader overview
as a first sten than a well founded special information. Nevertheless we believe to have obtained
beneficial knowledges.
The fabrik material, just like gfrp, demonstrated only a trifling damage sensitivity if cyclicly
tension loaded, although the stated stresses are related to the gross section, since a net section is
hard to identify in contrast to the presence of a drilled hole or a saw cut. The two solitary residual
strength data of damaged samples rise a little over the estimat.:d fatigue curve.
The corresponding undamaged lay-up made - UD laminae (0/90)3s proved to be rather stronger, even
takinq into account the different fibre content. Also a remarkable difference in the behaviour of
damaged and sound samples was observed. A 0/900 lay-up is notch sensitive when tension loaded. The
available data are not very easy to interprete, but the estimated fatigue curves of the damaged and
intact specimens seem to intersect at about 10 to 100 million cycles. The fatigue strength and the
residual strength of the damaged specimens seem not to have been affected significantly by the preceeding
dynamic load, whereas the fatigue curve for the undamaged test pieces decreases with increasing number
of survived load cycles. Not so the residual strength: again it reaches at least the range of the static
4$ strength obviously independent on the preceeding load history.
The multidirectional laminate (0 2;+/-4b/0 2 /+/-45/g0)s did not show such a substantial variation with
damaged and sound samples loaded in cyclic tension. The estimated fatigue curves decrease moderately
with increasing lifetime. The difference may be assigned solely to th.e different effective net section
caused by the indentation damages. The residual strengths are obviously independent on the preceeding
load history and reach the static strength or even exceed it just like the undamaged specimens.
Loaded incyclic compression, the multidirectional laminate behaves different. Incontrast to theI
I
~' ~ damaged spec.imens, the undamaged samples demoastrate a remarkable decrease of the fatigue curve with
increasing lifetime. Again the residual (compression) strength reaches the corresponding static value or
ven exceeds it as with the intact test pieces.
number of specimens damaged by impact ("dropped tool", and tested for fatigue life and residual
S2 , _h still is very small today. In general they seem to confirm the principal results given above
* F ndentation damaged samples (Fig. 14). Since the cyclic compression-tension load (R = -1) is one
) of the most severe harmonic load conditions lor the investigated multidirectional laminate both with sound
.I as well as with notched samples as reported by SCHUETZ, GERHARZ et. al, f5, 9) , we employed this type of
"loading with the impact damaged specimens. In contrast to the moderate influence of the impact damage on
the static tension strength the quasi-static and dynamic tension-compression strength related to the
"gross section drops drasticly. Neglecting the compression bearing capacity of the damaged area detected
by ultrasonic scan, we may estimate an effective net section of about 40 %of the gross section (Fig. 16).
12-6
That means that the effective net section strength (at R -1) is placed near the lower bound of the
compre3sion fatigue data (R -- ) of the undamaged opecimens. Concerning the damage propagation during
fatigue loading the various composites demonstrated significant differences. The woven fabric laminates
usually broke down suddenly by a crack transverse to the load direction crossing the defect, if present.
At the 0/900 specimens the edge corners of the upper 00 layer split off in the course of the dynamic
loading, caused by prevention of the transverse contraction by the 900 layer. The abrupt crack perpendicular
to the load direction ran through the indent defect. Besides, at damaged specimens a partial delamination
and splitting of the damaged outmost sheet with cracks parallel to the fibre direction starting at impact
45 9
locus was noticed. With the multidirectional laminate (02 /+/ - 4 5 /02/+/- / 0)s this splitting occured
earlier and always, even in the static case. Cracks and delamination ran up into the clamping attachments.
Often the 450 layer beneath the outmost layer was destroyed completely. Also as with the 0/900 laminates
the edge corners split off.
The damage propagation in impact damaged samples with multidirectional lay-up was not very
pronounced during the tension-compression fatigue loading (R = -1). No damage propagation was noticed
visually until quick break-down of the specimens, although an impressive motion of the damaged outmost
layers could be seen during load cycling. The ultrasonic scan detects only small changes (Fig. 16). The
4 damages introduced by the "dropped tool" probably are reflected by the fracture feature of the broken
test pieces. Sometimes, it is possible to distinguish fibre breaks at centre of impact frcm the surrounding
delamination area. For comparison Fig. 17 shows the fracture of an impact damaged, but staticly tension-
loaded specimen and demonstrates the faint detectability of the impact demage in this case, indicating
a different fracture mechanism.
Furthermore, the multidirectional lay-up presented a behaviour rather surprising at the first
glimpse. All test pieces, except the very short-lived ones, split in the midplane starting from the edges.
Fig. 18 evidences midplane delaminations registered by an ultrasonic equipment exploiting the reflected
signal intensity in a damaged ard in an undamaged spzclmen. Note that in each left graph the margin of
the scan is caused by the sensor size and does not indicate delamination. The upper sketch in Fig. 18
illustrates the remaining thickness which is free of delaminations, detected from the top side of an
impact damaged and tension fatigue loaded specimen. This behaviour evinced by the multidirectional
orientated laminate, not desirable in our opinien, can be prevented by a 0,1 mm thin glass cloth wrapped
around the edges a- manifested in Fig. 19. The covered edge exhibited no cracks or only cracks parallcl/
transverse to the fibre direction in the +/-450 layers not visible at the flank, whereas the unprotected
4! edge showed midplane delamination up to the centre line when loaded identically. Unfortunately, there has
been no possibility till now to investigate the influence of the wrapper upon the lifetime of damaged and
intact specimens.
{i CONCLUSION
HASHIN & ROTEM (6) have shown that in unidirectional cfrp the fatigue tension strength parallel to the
fibre direction depends only scantly on the number of cycles, and that the fatigue shear strength trans-
verse to fibre direction and the fatigue shear strength parallel/transverse to the fibres depend strongly
on the number of cycles or lifetime. That means, in the course of a load history in an ordinary multi-
directionally orientated, cyclicly loaded laminate the shear strength becomes more and nore the weakest
link. Regarding the damage tolerance of composite materials with respect to fatigue strength, the
. consequence is manifested in our experimental results. A laminate composition which enables to bypass a
damaged region with restricted load-bearing capability by means of longitudinal fibre stressing is superior
to any material which has to bear shear stresses principally via matrix or fibre/matrix interface.
Thus, the 0/90 lay-up is notch or damage sensitive when statically or dynamically tension loaded.
' The material is not capable to by-pass a significant distortion. The destroyed cross section principally
expands along the main axis leaving the rest nearly unaffected.
41:
12-7
Obviously, the inherent distortions within the fabric laminate are nearly as efficacious as the
artificial indentation defect. The tiny difference in general may be accounted to the different effective
cross section at the damage location.
The multidirectionally orientated laminate - considering the tension load case first - is definitely
notch insensitive on account of the inherent capability to endure inplane shear stresses by carrying the
load with help of fibres. Of course, microscopic shear stresses exist, too, which finally lead together
with the decreasing shear strength to the breakage of the specimen. We assume this mechanism to be right
with compression loads, too. On the other hand, the microsco, ic shear stresses are much larger, depending
e.g. on the filament alignment which results in a microscopic buckling stress state even in absence of
macroscopic buckling. The scanty test data may allow the conclusion that the exhibited reduction 3f
compression strength in sound samples in general is due to this microscopic shear stress which depends
substantially on the manufacturing piocess and perhaps the test method. In a real structure where
macroscopic buckling usually is not impeded as with our test pieces, the effective compression strength
is expected to be even smaller. Our first results with impact damaged samples - in this case an area of
about 10 cm2 is affected mainly by delamination and partially fibre/matrix debonding but also by fibre
fracture - indicate that the damage' .- ea can bear tension loads quite well, but minute compression
loads. The influence of the local b~nding stiffness reduction on panel buckling capacity in a real airframe
is expected to be more significant with impact damages than with indentation damages.
For 311 laminates composed of unidirectional sheets the decreasing fatigue curve in contrast to the
residual strength nf the not damaged samples indicates that fracture propagates rather quickly after
initiatic. by a small in-built defect.
Summarizing, the existing test results suggest that carbon fibre reinforced plastic generally seems
not to be damage sensitive in long-life structures (cyclicly loaded with low load level) since the fatigue
curves for damaged and inaffected specimens intersect at high numbers of cycles. An information pertaining
to the capability to endure solitary peak loads (in random load history) can not be derived from the
given data yet.
They indicate that a damage tolerant design requires a laminate composed of 00 layers in the principal
load direction, +/-450 layers achieving shear bearing capability, and 900 layers preventing excessive
inplane shear deformation by transverse contraction.
I Furtheron, the results support the opinion that usual theoretical tools in fracture mechanics are not
suitable to treat the time-dependent behaviour of real damages or even geometrically idealized
irregularities in many composites since they do not consider single sheets dedicated to different stress
f ill
REFERENCES
*_7
12-8
, t 2.7 mnu
fIf
p _ _ _ _ _ 12-9
2V
Zi
Z
-IU
,- R o0
-A- Co
WE . .N - I I C r
C2 R
A< M
g4 U.C
_q .a C C S2~
L
,=-=N
____ ___lea
.- I_ I '
C-' C
12-10
E-glass unid.
20 00 Co d e 69 , .. ,c.r.r.e
c. . .
N with respect to
m-m/,. large deflection
1500 -.
,- '~--,not
corrected
t "N
IN
U)N
I 1000
_"500-
0 10 20 /h30 40
40
30 10 20 304
E-glass unid.
10 Code 69
0
EL
0 10 20 Lh30 --- 40
.. w .
Fig. 2 Apparent Modulus of Elasticity and Ultimate Bending Strength of Three-Point-Bending Beams
12-I1
30
1
Jv DIRECTION
LOAD
~20 30
I5
ABSORBED
ENERGY
ANDULTIMATE
ELASTIC ENERGY
INAVOLUME
ELEMENT
LOAD
DIRECTION
30 1
DIRECTION
LOAD 25-
15 -1
1.1 RATIO
ofI l TOURE
RATIO 1.3 RATIOof MIXTURE
ofMIXTURE 3:1 ofMIXTURE
RATIO I iEXTUREI1:3 RATIOof
1:1 RATIO IIXTURE 3:1
ALTERNATING
STACKING
SEQUENCE ALTERNATING
STACKING
SEQUENCE
ABSORBED
ENERGY
INCHARPYIMPACT [UTAE ELASTIC
ENERGY
INAVOLUME
OF DIFFc.
SPEC IMIENS 1T FIBRES ELEIENT OFCHARN" SPECIMENS
IMPACT
DIRCTIOi
LOAO
30 -. 25-
S25 LOAD
DIRECTION+
15
f
W wekrIbimiI SA reNDWICHED
LAY- P rI lanSAllDWICHO LAY-UP
wekvfiimioile
stinerfireIssipeuokir fibre Inmkiplaso Isk.I.~lla
stroo fber
RATIO
OFMIXTURE 1-2-1 RAI FMIXTURE 12-1
ABSORBED
ENERGY
INCHARPY
IMPACT ULTIMATE
ELASTIC INAVOLUME
ENERGY
SPECIMENS
OFDIFFERENT
FIBRES ELEMENT
OFCHAIIP
IMPACT
SPECIMENS
Fig. 3 Totally Absorbed Energy and Elastically Endured Energy in Charpy Impact Beams with Various
It. Unidirectional Reinfrrcad Fibre Compositions
].(, S1
12-12
TIME- I
0 00 0 :FIX.CLAMPW/'
/
SPECIMEN
ELECTRO BARREL .. iD _
S....../ 't
I~x
CLEO # U -'~
I(...LkEE [TISPLACvEMEN5
VOLT*TER
11 T' ACTION
000 ELECTRO-OPTIC FS
ADDER RI6HTANOLEFT
SABOT
SOF PENDULUM
;jN
Aq
""ee m 1 a =2
(j". - - - -- - .
12,1
IVx
I covered
_311
/ ~rotected
v 34,3 rn/s
-A
Nit
-t: A i5
12-15
10.
f
"T
3-~
Fig. 8 Principles of influence of Impactor 'elocity and Support Conditions on Specimen Response
Above: Deformation in Large Plat.s Depending on Impactor Velocity (Evaluation Fig. 7)
Below: Response of GFRP Fabric Laminate on a Dropped Tool (m = 528 g) at Narrow and
Wide Support at v = 5.9 m/s
12-16
4 1
j I
!
. 1.
Fig. 9 Demonstration of Impact Resistance of GFRP versus CFRP with the Leading Edge of ALPHA-JET
Horizontal Stabilizer (Top)
Load: Impact by a Glass Sphere 21 mm Diameter at v = 54 m/s
Material:, 3 mm SC130/Code69 0/900, covered with 0.2 mm Aluminium Foil (Middle)
0.8 mm GradeO02/ Code 69 Sandwiched by 0.9 mm E-Glass Style 181 (Bottom)
(>
12-17
At
:k
I
J ~1
IO
Fig. 10 Defects Introduced in GFRP and CFRP Simulating the Response of Oropped Tools
Above: Threesquare Indent in CFRP, Die Force 5 kN (left)
Glass Sphere Impact on GFRP Fabric 1.7 mm, m r 13 g, v = 45 m/s (right)
Middle: Cross Section of Glass Sphere impact on GFRP Fabric (above)
Cross Section of Threesquare Indent in GFRP, Die Force 2.F kN (left) 5 kN (right)
Bottom: Cross Section of Threesquare Indents in CFRP T300/Code69
12-18
GD AC
40 N
Iq Ii !
00
-I -0 m a
'1 ~ ~ ~, Fig.
C3 -I Faiu uremfGFPFbi
12-19j
2111012
NZ 200 -
S M239 -. 'I
S0 100 200
Usm N
prupigitlin of dellaminutlln
&
t f-25.H
At I Mcm IN -
C400
11. 07-~~~~~~ ti26fl~ N'SVIL N ~4 6
10
sm~s~g
2w2.N~,
~titmsch
.'~10 pukij0. 10
I~~ 17______
VAITO _ _ _ _
ype ofFalur
Fig.12 Mdc
_
OFYL
_
FAIUR
nd amae Popagtio
_ _ _
inGF~Ia
MODE1
mis i
sicLminae
Ai
VI
'
[
sttsh2
0 X5 &4 n
5.,.6m Pilc
sieiafit 5'
A*VRAINOFFIUEMD
12-20
,ooo
N/mm'
rioo
x1000BO
F-SNz 1000
R00T-
Io"3
5-00
T30O/C03E 0
..-
790 .- o
700 "00'
-600 6400
cc 500
S30000
T300/f;OPE59
IOZ/1A5121/'451Hs T300/COOE69
700 ....
I500 0 - 590 o .
20 tt
en300 ~300 W oiC
;'.0
1 o
OAIAGED 0i
N BAMAK
''"oF
NUMBER OF CYCLES
-nR DAMAGE TOLERANCE
I
I ,Fw.,
NUIBER OF CYCLES
DAMAiZ TOL .AiCE
N
Fig. 13 Fatigue Curves and Residual Strengths (in Tension) of CFRP Specimen with Indentation Damages
14
12.21A
T300ICode69
lg
I.-
368 r4Smm
to
208
110
23S
NUMBER OF CYCLES N
A ~~~~~~Fig.
14 Fatigue Data of CFRP Specimens with ~DVRI DMS OEAC
Impact Damages ('Dropped Tool')
tabs
z
.0.~n~n1.2
LLU
01
600
~L'
U.U
f If
As 3
Fiq. 16 Details of Damaged and Fatigue Tested Specimen Made of T300 /Code 69 (02 /-45/02/-45/90)s
Impact Parameters: m = 244 g, v =6.8 m/s, Sontact Tip Radius = 5 mm,. Eki 5.64 J
Plate Clamped in 130 mmnDiameter, Irrev. Absorbed .nergy Eab 3. J
Load Parameters:, G:max = 103 n/in 2 (R =-1), N = 1.3 Mio Cycles
Residual Compression Strength 6r 216 N/mml2
;2:2e
12-23
A'
Fig. 17 Fractures of Impact Damaged Specimens after Cyclic Fatigue Loading (R = -1) (above, middle)
I-"
not damaged
damaged by
threesquare
indentation
I T3OOICode 69
il 1021/t45/02/45/90 Is
N O N=2,8 105
IL I= 2.5mm
10=317 pin
R =.01I
5o=458 N/mm 2 N= 0
5
-j
N=2,8"10 5
*DFVLRDAMAGE TOLERANCE
C-Scon-frontside
--
' w' =
: ' iDFvL' DAMAGE TOLERANCE
mE = 3 -
12-25
corner
fspi.
~V
I0
4-
4-
C,
Mesa-
Fig. 19 Prevention of Edge Effect in Tension Loaded Coupon Specimens by Wrapping Around a Thin
Glass Cloth; Material: T.300 /Code69 (02/t45/02/ 45/90)s
Load 'Conditions: G 600 N/i, R = .01, N 1.1- 10~
R3-1
RECOR)ER'S REPORT
by
K.Brunsch
Messerschinitt-B161kow-Blolhn GmbH/UF
Postfach 80 11 60
8000 Milnchen 80
West Germany
The paper by TorTes presented information from service experience and from tests with rotor blades, that is with
complete FRC structures. The remaining papers present somewhat more basic research aiming to relate impact defects
with residual strength, static and fatigue, and also fracture toughness with residual strength after impact.
Torres very favourably highlighted the excellent structural properties for FRC for rotary vi'ig application.
Information on residual strength after extensive damage caused by ballistic impact is given. 0, .he other hand, the need
for sufficient erosion protection was also pointed out. The presentation end. ' % ith a film of impact tests of SA 340
main rotor blades against wcoden rods. Seeing the blades cut a 200 mm thick rod, the recorder was impressed and
concerned. The concern is on v,n.-ther such information might reduce pilots' carefulness.
The paper by Dorey presents the results of impact tests and of tests with machine notched specimens. He corrects
"forthe damage zone (rp) and by that means achieves reasonably good prediction of the residual strength of notched
laminates. Kc is between 20 and 50 MN m-3 2 , which means that considerable notch sensitivity for some laminates was
found. Good correlation was also found between Charpy impact energy and fracture toughness. The reduction of
residual strength versus incident energy appears to be more troublesome to predict. The same energy may cause
differences in residual strength of up to a factor of 2, depending on impact velocity. The information given is that
specimens of orne particular laminate failed at incident energy loads of 3 to 8 J wheor loaded to about 40% of tensile
ultimate strength.
The paper by Card and Rhodes presents test results of 560 specimens and components ihat
being compression loaded. The research work reported was done with sandwich beams, laminatedhad been impacted
plates and top-hatwhile
.stiffened panels. The most important conclusion from this work is that, with the material inmestigated, there is a .003
compression strain threshold, i.e. components compression loaded to this strain do not fail whin impacted with up to
15 J energy projectiles. This strain threshold was found to be valid when using a resin system with .01 ultimate strain.
A tougher matrix system - about .04 ultimate strain - considerably improves damage tolerance. Taking into account the
well known rules of stress and strain magnification in transversely loaded composites a .01 ultimate strain matrix system
does appear to be too brittle. Even with that matrix system and the low .003 strain 'hreshold, a composite wing is weight
effective in comparison to an aluminium alloy wing.
The last paper of the session by Aoki an-a Stellbrink reported many test results, the most important ones covering
the influence of defects on the fatigue stmngth of CFC specimens. Prior to fatigue testing, the specimen had been
damaged by ball impression, three-square indentation and ball impact. Fatigue tests were done with R = .01,
R = -0.01, R=: -l and the residual strength after a given number of cycles was determined. Based on a limited
number of specimens, the conclusion can be drawn that high-cycle fatigue strength is reduced much less than static
strern -th. Toughness of resins, which contributes to shear strength, again was found to improve damage tolerance.
"P I Generally it must be ;aid that, apart from the excellent papers given, there remains a considerable amount of work
0 to be done with reference to damage tolerance. For example, the influence of moisture and of ti.e load history prior to
, an impact should also be known.
13-1
"-7F
EVALUAT ION DU COMPORTEMENT A LA FOUDRE
DE STRUCTURES EN MATERIAUX COMPOSITES H4AUT MODULE
par
J. ROUCHON, D. GALL
CENTRE D'ESSAIS AERONAUTIQUE DE TOULOUSE
23, avenue Heairi Guillaumet
31056 - TOULOUSE CEDEX
France
RESUME
Le faudroiement en vol d'un adronef est un phdnomane rolativement frequent qui se traduit par des en-
dommagements voire des destructions au niveau, d'une part des 6quipements, d'autre part des structures et
en particulier celles eilmatdriaux composites.
Ce document, apr~s quelques g~ndralit~s sur ls fcudre, (mesure dea caractdristiques de is foudre en
vol, proc~d~s de simulation au sol), pr~sente des r~sultats d'essais effectu~s en laboratoire sur des
fchaiprillons (monolithiques en carbone-6poxy, sarndwichs avec revktements bore-6poxy), des 6l6ments de
structures r~els, ainsi que lea proc~dds de contr6le aasocigs.
On estise que globalement, il existe dans le monde A tout instant 2000 a 5000 orages produisant envirc'n
une centaine d'6clairs par seconde. Ces orages ne Sont pas r~partis unifornidment autour du globe terrestre.
On note en effet environ 200 jours d'orage par an dans les zones 6quatoriales, alors que dana lea zones
temp~r~es on n'en relhve seulement que quelques dizaines.
On comprend d~s lors que les agronefs, 6voluant A proximit6 ou dana lea masses nuageuses, sojent fou-
'1 ~ droy~s en vol.
Les atatistiques, A l'dchelon mondial nmontrent qu'5 l'heure acutelle, la frdquence des foudroiements
eat de l'ordre de
- otsla20 ersd o ou e orse oescures
- 1 toutes les 2000 heures de vol pour lea courts emonscourriers.
Cette diff~renee est due au fait que les aviona couirts et moyens courriera ont un temps de vol par
mission en basse et moyenne altitude supdrieur A celui des longs courriers. L~emaximum des foudroier'ents
a lieu A une altitude comprise entre 1000 et 4000 m
Il importe donc, qu'au cours de Ia trentaine de coups de foudre que subissent lea avions pendant leur
2 - ECAIIISMES DI!FOUDROIEMENT
Z(KM) C
1+ +* +~
1 + +4-
60 + + -415
902
4 . -
2+''' ~*ai1 +5
01
Ql:+20C A 2KM
Q2:-42)C A 5KM
Q3:+40C A 10KM
D'UN NNUAGE
ORAGEUX
it
'~'SCHEMA
~fig. I
13-2
L'origine de la foudre se situe g~ngralement au nive..u des nuages A d~veloppement vertical tels que*
les cumulo-nimbus (fig. 1). A l'intdrieur de ce type de nuages, les violents courants d'air entralnent
les gauttelettes d'eau qui par frottement, se chargent en 6lectricitg. Ces gouttelettes se r~partissent,
A l'int~rieur du nuage, en Tiots de charges glectriques plus au mains nombreux.
lorsque le champ glectrique E - VA', V difference de potentiel exprim~e en volts (plusieurs centaines
de millions de volts dans le cas de foudre), d la distance entre les !lots, atteint une valeur suffisan..e
(environ 500 000 V/rn) un arc glectrique jaillit entre ces charges.
- les coups pasitifs amorqage des charges positives du nuage (10 % des cas)
- les coups n~gatifs amorgage des charges negatives du nuage (90 % des cas)
L'arc 6lectrique qui jaillit entre les charges 6lectriques de signe oppas6 eat canstitu6 de deux types
de composantes
} -
-
des composantes impulsionnelles
des composantes persistantes
}A A: Cowesnpsrtes impulsionneiles
B: Courants estas
-fig. 2-
j. ~ L en~a~ed ': :~iud:dat~::lu.c ulusdzie ecetd'nio
L'ensemble du phdnomaie a une dur~e qui peut varier entre quelques dizaines de mil~lisecondes et 1,5 a.
20 00A3 0oK
di
yi Id
I max
TO Ti t
-fig. 3 -
13.3
( 7!-. , __ __ __
-fig. 4 -
- Ia dur~e To
- la valeur max, I max-
- le transfert de charge Q )idt
Le tableau suivant donne I titre indicatif, une valeur mcyenne et une valeur maxi de ce3 difffirents
paraim~tres pour les coups de foudre entre nuage et sol.
VALEURS VALEURS
I UNITES MOYENNES MAXIMALES
COMPOSANTES IMPULSIONNELLES
*Vai'qux maximale Imax kA 40 200
I
I
* Valeur minimale
~fig. 5
2.3. Foudroiement d'un avion en vol (fig. 6)
o:::ls::c::r::tit
vio et fudoy6envol, sastructure esc travers&c par le cou::r:ant
Lorqu'n V
Sr crtanespariesde ~avonle point d'attachement de l'arc est
~fiiesau
tiqes nt t62..
fixe Vaor stquesur s lespace, alors que l'avion, lui est mobile (fig. 7).
4~ I A?_ __ _ __ _
13-4
fig. 6-
2 2,
~TI,
ffig. 7 -
Sur certaines parties de la structuire, ce "balayage de foudre" se traduit par une succession de points
I Ii
d'attachement .On d6finit ainsi 3 zones sur un avion
zot.a I surface pour laquelle il y a une grande probabiliL6 qu'iI se produise le premier impact de
-zone 2
foudre
surface pour laquelle il y a une grande probabilit-6 qu'il y ait "balayage de foudre".
ceptibles d'atre parcourues par conduction par le courant foudre entre deux points d'atta-
chmet
13-5
-fig. 8-
3 -CONSEQUENCES DUiFOUDROIEMENT
Les cons~quences d'un foudrajement sont de deux types,celles li~es aux effets directs et celles li~es
aux effets indirects.
Les effets directs sont ceux occas.& idsA la structure et qui se traduisent par des d~g~ts mdcaniques
points de fusion, 6ciatement de carane, etc.
Ces d~gats ddpendent bien entendu des caract~ristiques 6lectriques du courant foudre, cependant sevls
certains paramatres jouent un r~i, pr~ponddrant
L'attdnuatiun des champs glectromagn~tiques apport~e par effet de blindage de la peau de I'avion joue
alars un r~le tr~s important.
4 - EXPERIENCE EN VOL - MESURE DES CARACTERISTIQUES DE LA FOUDF.E A L'AIDE D'L'W AVION TRANSALL
Les caractdristiques de la foudre donn~es au paragraphe 2.2. proviennent de mesures au sal, c'est-A-dire
4pour les coups de fouarq entre nuage et sal. On ne connait -1l'heure actuell'i, aucune valeur des caractfiis-
A ti~ques des coups de foudre entre nuages ou a l'int~rieur d'un nuage, coups de foudre deux fois plus nombreux
que ceux impactant le sal.
Afin de connattre las caract~ristiques des coups de foudra en altitude, lea Se~vices Tw~hniques
AMronautiques frangais ant ddcid6 de lancer un programme de mesurea.Le Centre d'Essais en Vol (CEV) et le
Centre d'Essais Agronautique de Toulouse (CEAT) avec Ia participa,.ion de l'AIA de CLERIAONT-FERRAND at de la
SEFTIM ant, au cours de l'6tg 1978 rgalis6 une campagne de mesures a l'aide d'un avian TRANSALL (fig. 9).
L'avion a 6t6 6quip6 de 2 perches de 4 m. de langueur (fig. 10) munies chacune '1'un shunt de mesura coaxial.
13-6
fig. 9
Outre ces mesures sp~cifiques, d'autres param~tres ant ft6 etiregistr~s (champ magn~tique, courant de
peau, surtension dans les r~seaux de bard et sur 6quipements).
I -fig. 11-
,'
13.7f
-fig. 12-
Au cours de 17 vols, 13 enregistrements ant W effectu~s. Les r~sultats obtenus montrent que lea va-
leurs du courant mesur6 restent a l'intgrieur de l'enveloppe des couranta mesur~s entre nuage et sol.
K~ -- , -1
:4n
fig 14 -
V1
., 1
- fg.74
Fp,
13-8
Un deuxi~me simulatcur (fig'.15) g~ngrateur dL choc de courant perxnet d'obtenir des impulsions de courant
(installation du CEAT, 200 OOOA suivis d'une composant.e continue de 500 Coulomb).
-fig. 15-
Le courant obtenu, inject6 aux points d'impact pr~alablement d~termin6s en choc de tension, permet
d'6valuer la tenue des structures. CVest en effet le courant qui provoque. lea endommagements sur les mat6-
riaux composites.
Le phdnom~ne de balayage de foudre peut tre r~alis6, soit par d~placement (lineaire ou circulaire) de
1'6prouvette, soit par soufflage de l'arc (soufflage magn~tique ou soufflage d'air).
Apriis une mise en Evidence de la vulnirabilit6 des mat~riaux: composites vis a vis de la foudre, les
ob ectifs des r~sultats pr~sent~s dans cette conf~rence sont les suivants
Si dana un premier temps, les essais ont fit r~alis~s sur des composites carbone et bore-6poxy, les
p exp~rimentations actuelles ne concernent plus que le carbone-6poxy.
2i
I"
La difficult6 d'o~tenir expfirimentalement un transfert de charge glectrique avec le bore-6poxy (non
conducteur), ne permet de comparer lea deux mat~riaux que via a vis d'un choc de courant seul (composante
impulsionnelle).
Les deux photos ci-apr~s montrent lea e~gats obtenus sur deux plaques fl'Apaisseur fiquivalente (0 mm)
apras un choc de 200 kA sans transfert de charge.
carbone-dpoxy bore-fpoxy
13-9
6.2. Influence du mat~riau constituant l'ttme en nid d'abeille des structures sandwiches.
Les r~sultats ci-apr~s ont 6t obtenus avec des revatements en composite bore-Epoxy d'6paisseur I mmn,
et une ame en nid d'abeille m~tallique ou Nomex.
Montage d'essai:
AG
f3
13-10
Avec le nid d'abeille NOMEX non conducteur, l'Sme ne paralt que tras peu endomsmag~e, la perforation du
revatement est 6galement moins importante.
Apr~s quelques essais priliminaires r~alisgs en France sur ces difffirents proc~d~s, le systame des
grilles de bronze nous eat apparu comme ayant le weilleur rapport masse surfacique/efficacitd.
Une campagne d'essai r~alisge dans le cadre d'un programme de coopgration avec la EPA (I.A.B.G.) nous a
permis d'6tudier plus pricisfiment Ilefficaciti de ce type de protection vis A vis des difffirents param6Lres
de l'onde de foudre (choc de courant et transfert de charge).
-matgriau exp~riment6 carbone 6poxy T300-N5208, en plaques monolithiques 8 plis orthogonaux (0,90*)
-Protections 7 L2 2 2
grilles de bronze de masses surfaciques 47 g/m -80 g/m et 250 g/m (placges sur une seule face,
c8t6 filectrode).
Les illustrations ci-apras permettent de comparer l'efficacitg de ces protections vis a vis de trois
types de foudroiement.
- choc de 200 kA sans transfert de charge
- choc de 200 kA avec transfert de 210 Coulomb
~ ~- Choc de 50 kA avec transfert de 50 Coulomb
F
so1 rne
il
vA
rleboz
ae
Prtcin rleboz
ae 28M
ci 1t6 6Iecrod
op8os9e 596
clc:01 a~iwabn-px 30 50
13-11
sons protection avec grille bronze ovec grille bronze avec grille bron~ze
electrode
N8982 986
me MB989- -N48 39 L~~
trainie
2 0C 50m li 0,9' yn~rqe
face
________Me 995 9B
letrnsertd'eege(orcoc 0k 0C
LMI4
4 pr ine erfraton plu ou_momso~aroe-x
iTp30ant deNete2laue
t .0m
Viueleet, laspc ec omaeettu f8i copaale 0'
c901qu syerait roolupa e ho
dutobepari solliciotios
des derefaige tierletti donclusions sune gandespri ecteprelraed
- Laproit d etoWaingarieu
u pstpopruturnele pro le mset dourciu
d desaoirgsille
eisteu dommages suppl&
_ _ _ _ Viuelmet
_ _ _ _ _ _ de_ opral eu uisri rvq-. a eco
_ _ _ __eettotAfi
ias vies du rjetl onoii
13-12
Quatre chocs identiques ont 6t6 rfalises sur des plaques monolithiques 8 plis, de 400 x 400 Tam,avec le
mame montage d'essai que ci-4essus (cf. 6.3.). Deux 6prouvettes de traction de 200 x 25 m ont pu 9tre ainsi
prglevdes au::our de chacun des impacts. 290
- 265
110
w!
00
0 k,'P Ip
., L" . , I
d
1'
g/m2
t.....L.. .
La photo ci-apr~s illustre le principe du pr~l~vement des 6prouvettes par rapport aux impacts. Soit au
~voisinage immddiat de Ia perforation, dans une zone gdn~ralement noircie suverficiellement mais 6Vun apparait
2
en surface ni fibre coup~e, ni risine bralde.
- 50kFu de
-0Cae-ri-l rone d 80g/
TI
Quelle quc soit la configuration du courant de foudre, et le type de protection, les rfisultats de ces
essais mcaniques Wuont pas mis -en vidence un endummagement de la risistance autre que celui propre A la
perforation.
1 13.13
6.5. Approche exp~rjmentale du phinom~ne rgel-foudroiement avec balayage.
Les experimentations en vol effectufies sur TRANSALL ant permis de mettre en Evidence le d~placement de
l'inmpact de Ia foudre sur la structure de l'avjon.
Cet eff~it de balayage impliquant un transfert local d'Energie moindre que dans une excpirimentation
statique, les .ionages provoqu~s seront de moindre importance.
Dans le but d'une simulation au sol plus fidale, le C.E.A.T. a donc d6veloppi un dispositif expgrimental
otale balayage est obtenu par rotation, devant 1l'ilectrode, d'un disque constituant l'6prouvette a essayer.
46
Dans ce principe, Ia vitesse de balayage eat vite Iimit~e par les forces centrifuges et le diam~tre
maximum "raisonnable" de l'6prouvette.
Des essais ant pu 6tre rdalis~s sur des disques en mat~riau composite de diamatre 600 mr avec une
vitesse de 15 km/h au niveau de l'6lectrode.
Dans ces conditions expdrimentales, des r~sultats obtenus aur disques sandwiches avec ThneNOMEX et peaux
en, carbone-Epcxy T300-N5208 8 plis, ant ddmontr6 l'6quivalence des donanages entre lea chocs suivants
- 200 kA - 150 C pendant 0,3 s. de halayage
I
-200 kA - 50 L.sans balaysge (statique)
Ces essais de foudrojeme-nt dynamique ant 6galement permis de mettre en Evidence l'influence de Ia couche
I de peinture protectrice.
h " Las photos ci-dessous montrent l'effet de concentration des ddggkts provc'qufi par una peintura non conductrice.
Configuration d'essai 200 kA plus 1'50
C pendant 0,3 s. avec balayaga 15 km/h.
4enepit non
13-14
L 7
~
- ESSAIS DE FOUDROIEMENT SUR STRUCTURES REELLES7
Leur objectif eat la determination des points d'impact et des cheminements de la foudre stir la structure.
Ils permettent donc d'ivaluer l'efficacit6 des paratonnerres "naturels" constitu~s par les ferrurka. - en-
cadrements mdtalliques de la structure, ou art.f~ciels (bandes conductrices disposdes sp~cialement pour .apter
et Geouler le courant de foudre).
Ces essais r~alis~s sur des installations de choc de tension nficessitent des Evaluations compl~mentaires
en choc de courant afTin de quantifier les risques de donmmage rfels dans lea zones concernges par l'iiupact et
le cheminement.
L~edocument ci-apr~a illustre tineassai d'efficacit6 de bande paratonnerre segmentfie stirun rad~me en verre-
Epoxy.
La bande sur laquello on distingue tr~s bien l'6coulement de Ia foudre Eitait ccinstitufie de pastilles canduc-
trices de 02,5 mm et distantes de 0,3 mm (fixation sur film do r~sine Epoxy suivant principe des circuits --i
prim~s).
Porte-antenne
mrt-alli tieU
Isolateurss
8 -CONCLUSION
Lea premiers probl~mes de vuln~rabilitE des matfiriatix composites A la fuudre sont uppartis avec lea rad~mes
et divers cargnages d'antennes en stratifi6 verre-6poxy. Ces probl~mes istaient at demeurent d'au':ant plus
complexes A rdsoudre que lestr solution va toujours a l'encontre des impdratifs de perm~abilitEi magn~tique pro-
* pres a ce type de structure.
Au d~otit des ann~es 70, pour la premiare g~n~ration de pi~ces structurales en mat~riatix composites haut
*1 nodule (gouvernes et caissons d'empannages) sans impdrarif de perm~abilitg miagn~tique, des syst~mes de protec-
tion simples tels que des bandes conductrices en clinquant ont 6tg syst~matiquement prdvus.
* ~Vu la surface relativement petite da ces structures, leur position sur l'avion et l'importance des enca-
drements mitalliques, la probabilit6 d'un impact de foudre direct stirle composite eat tras faible. Ces protec-
tions lourdes se sont donc av~rdes rapidement superflues.
L'6volution actuelle des &.pplications vera des structures beaucoup plus importantes (dfirives, voiluras) avec
beaucoup momns de parties m~talliques remet le problame de la sansibilit6 A la foudre aui"goat du jour".
Totitefois, atixdommages structuraux directs s'a~joute maintenant le problame de l'ouverture difilectrique crEE
par ces 6l9ments, at lea risques de destruction des circuits Eilectriques situ~s I l'int~rieur qui en d~coulent.
Dana l'avenir, D~ eat mame probable qua c'ast sous ce deuxiame aspect qua se posera assantiellemant le problime
* ~du foudroiement des structtires en matEriaux composites.
contra6inemefoaniqnupa
A cocmcaique,
a np~utVmpeur onsaid re consientrabedunrite par l'utilisa-
tals dommages. Cependant, ii nous reate encore A acqu~rir tineexp~riefice compiEmantaire auiniveau'das liaisons,
LIGHTNING PROTECTION
CONSIDERATIONS FOR GRAPHITE/EPOXY
AIRCRAFT STRUCTURE
S. D. Schneider
THE BOEING COMPANY
Mail Stop 47-31
P.O. Box 3707
Seattle, Washington U.S.A. 98124
SUMMARY
When advanced composites such as graphite/epoxy were first being considered for aircraft structure, a common belief
was that lightning would heavily damage the structure. This belief has since proven to be false. Advanced coi-posites
react to lightning strikes in a manner different from aluminum, but the resultant damage is by no means alarming.
There are two basic types of lightning damage to which aircraft structures are subjected. One type occurs at the arc
attachment point, and the other occurs in the intervening structure between lightning attachment points. The effect
of graphite/epoxy structure on each of these phenomena is discussed.
Known, documented techniques and design philsophies for pratection against lightning-caused structural damage are
reviewed for classical aluminum aircraft tru.cture fbi cated with metal fasteners, and the impact of graphite/epoxy
on these classical approaches is addressed. Detailed lightning test criteria, test techniques, and criteria also are given
't and related to graphite structures.
INTRODUCTION
The advantages of graphite/epoxy structure are so overwhelming to the structural designer that its widespread use for
aircraft structure is inevitable. The graphite/epoxy ma rix is roughly 1,000 times less conductive than a comparable
aluminum structure. For this reason, use of the material causes some concern to the engineer involved in aircraft
structure lightning protection.
Lightning pratectior, of conventional riveted aluminum structure has been straightforward (ref. D). The conductivity
of the basic airframe serves as an adequate current path for lightning d.scharges. Protection problems created by
dielectric sections and access panels were easily corrected by providing conductive paths to the airframe.
Since epoxy matrix is dn excellent electrical dielectric, any current flow through the composite must occur in the
graphite fibers. Graphite fibers are sufficiently resistive that the localized flow of currents of several amperes can
cause localized degradation of the matrix and, in some cases, destroy the structural integrity of the composite.
A number of lightning protection concepts have been developed for graphite (ref s. 2, 3, and 4) and can be used effec-
tively where needed.
part inThe
1 graphite/epoxy the protection concepts
localized area generally
requiring involveMany
protection. adding a conductive
factors must belayer to the however,
evaluated, outside surface
before of
a the
lightning protection system can be selected. These factors include cost, weight, complexity, criticality of the part,
and predicted frequency of lightning strike to the part.
There are two basic types of lightning damage to which aircraft structure may be subjected. One type occurs at the
arc attachment point; the other occurs in the intervening structure between lightning attachment points. The effect
of these types of damage to graphite/epoxy must be considered when evaluating lightning protection requirements.
Additionally, the protective scneme4 must be consistent with expected level of lightning threat (which varies with
location on the aircraft) and the structural and economic penalty of not providing protection. Protection systems
developed must not only prevent degradation of struc~ural integrity, but must also be conistent with the aircraft's
Atotal environment and designed in such a manner that damage is detectable and repairable.
This paper only addresses protection against structural damage (i.e., structural degradation) due to lightning. Fuel
tank arc and spark protection are not discussed. (See refs. I through 4 for more information on these effects.)
As shown in the table, a lightning strike is typically 0.2 sec and can be as long as 2 sec. An aircraft can continue to fly
through a lightnin strike after attachment is made at a forward point, such as the nose. Therefore, aircraft surfaces
S.'T) aft of a forward lightning attachment point must be protected against "swept" lightning strike attachments.
V, -
14-2
PROBABLE
CHARACTERISTICS TYPICAL WORST CASE
The classical approach to aircraft protection has been to zone the aircraft acco,-dwg to the probability of lightning
strike attachment. The zones are defined below and illustrated in figure 2. Graphite/epoxy is sufficiently conductive
that these zones still apply to aircraft with minor oi major portions of graphite/epoxy structures.
* Zone IA: Initial attachment point with low piobability of flash hang-on, such as the nose; i.e., the lightning stroke
has a high probability of sweeping aft
e Zone IB: Initial att.,chment with high probability of flash hang-on, such as a tail cone
* Zone 2A: A swept stroke zone with low probabihity of flash hang-on, such as a wing midcl.ord
0 Zone 2B: A swept stroke zone with high probability of flash hang-on. such as wing trailing edge
Since there is an entrance and exit attachment point for each lightning stiike, intervening structure between the
attachment points must be designed to sustain the passage of lightning currents.
Structural damage or degradation due to lightning is generally created by two effects. One is thernal; the other is due
to high current aensities.
Temperature of the channe! (arc) of a lightning strike can be as high as 27,000 0 C, and aircraft surfaces contacted by
l'ghtning may be exposed to teml'eratures that high. Obviously, few materials can withstand these temperatures for
very long without thermal effects such as burning, charring, or vaporization. The length of time that lightning can
remain attached to an aircraft surface depends on the aircraft lightning zone. Probable worst case dwell times
(thermal exposure time) will vary according to the aircraft lightning strike zone. Attachment point times for the
various lightning zones are given in table 2.
High current density is also a damage mechanism caused 1-y thermal heating or explosive vaporization of the individual
graphite fibers. The tremendous peak current possible in lightning results in high current densities at the arc attach-
ment point. However, the currents tend to diverge rapidly in materials like graphite/epoxy laminates and aluminum,
thereby reducing the current densities at short distances from the attachment point.
, bank of capacitors foi a total capacitance of 454 IJF at 20,OOOV for the desired energy figure of merit (i.e., action
integral).
Various combinations of the four components (fig. 4) are used according to the aircrait lightning zone for which the
hardware under test is intended and are structured to be compatible with the dwell times shown in table 2. For
S- exampl,, zone lB uses all four components, because lightning can attach there and remain attached for the duration of
the lightning strike. Zone IA, however, uses only components A and B, since the strike will sweep aft after
"I
attaichment.
f!
14-3
ZONE 1A
COMPONENT STROKE)
A (INITIAL COMrONENT D (RESTRIKE)
PfAK AMPLITUDE-200 ktA 2 PEAKAMPLITIJDE-100kA 2
ACTIONINTEGRAL-2 - 10 A . ACTIONINTEGRAL-0.26 x 106A .uc4
COMPONENTB (INTE:RMEDIATE
o ~MAXIMUM
VEECNARGETRANSFER-I0C
CURRENT)
~ ACOMPONENT
AMP
LIT
C
(CONTEREIATEN
CURRENT)
-AA U
OE-20A
IM TIME
NOT TORSCALE-
5'
not degrade the Llectrical continuity of
the Faraday cage. Test specimens are rigidly installed in a nonconductive
phenolic block, ories ited such that lightning
currents attached to the upper one-third of the panel are forced down
througt. the lower two-thirds to the electrical
ground return on the lower 6-in (152.4 mam) end of the panel.
SZONEPANEL
{ ~INSULATORS II
.S
* O.l
5
.67 x 1O A 2 SI
i&06,A1C.
.l " " "04 ,10S
AF I
i re 7
MFigure
* L..-O.5xiOAZsKc Zone IA Lightning Test to
6. Zone 1,A
(Reduced Level) Lightning Test to 10 Ply Painted Graphite/Epoxy Test Panel
PaintedAluminum Test Panel
L
fT ~.nuisance
~ Thcse results suggest that aircraft locations subjected to zone J lightning attachments
damage if aluminum is used instead of graphite/epoxy. Because areas such as (extremities) will have less
S, structure, the benefit of using higher-strength materials like graphite/epoxy wingtips are usually secondary
is less obvious and aluminum f airings
;; should be considered.
14-4
Fiberglass and Kevlar also are receiving increased interest for structural applicdtions, especially in secondary or
lightly loaded structure. Classic protection techniques for fiberglass have included aluminum bus bars, buch as for
wingtip and radome protection (re!. 1). Fiberglass and Kevlar are nonconductive and, as a result, lightning attaches to
the aluminum bus bar (i.e., diverter strip) instead of puncturing the fiberglass or Kevlar part. Graphite/epoxy,
however, is so conductive that the divcrter strip approach does not work. That is, lightning sees no preference for tie
aluminum diverter strip and is just as apt to attach to the graphite surface. If protection is considered desirable to
reduce damage to a graphite/epoxy part, such as a wingtip, a protective conductive coating should be considered.
Candidate coatings are discussed in the next section.
An estimate of the damage to the A component of lightning is illustrated in figure 8, excerpted from reference 1. The
data in this figure must be considered approximate because they came from a variety of tests. The data approximate
damage as a function of the number of plies and the energy factor. The ordinate value of 2.0 on figure 8 corresponds
with a zone 1, or those areas that require 200,OOOA (see fig. 2). Similarly, the 0.25 level corresponds to the swept
stroke zones. As shown, widespread damage does not occur at either test level, except for very thin laminates. Thin
laminates typically are used only in noncritical areas (secondary structure). As a result, protection considerations
against structural damage caused by lightning are generally economic and not involved in flight safety.
Reference I has another clever way to approximate danage. The data used to compile figure 8 also were used to
2
produce table 3, whi'h shows the approximate action integral (energy factor, fi dt) that various cross-sectional areas
can sustain with minimal damage.
Total elimination of damage may not always be practical or desirable. Localized heating and thermal erosion will
occur on most any material, including aluminum, due to the high temperatures of the lightning arc. It is important,
however, to approx;mate the extent of damage by test or analysis. As detailed in reference 1, the information shown
in table 3 will provide an estimation of damage when used with the following equation (see fig. 9).
d>A
Ttr
2007
1.50
OAMAE DAMAGE
GENERETOARCTIO
SMFII GEEALDRETO
NUMIR OF .PLIES
Figure & Estimated Visible Damage to Unprotected Figure 9. Composite Thickness Versus Aroe of Darnag
GraphiteComposites From ConductedCurrents
Consider the following example. A graphite/epoxy laminate is proposed for use in a zone 2A, fuselage section. For
structural or repair reasons, it is desirable to limit the damage to an area with an approximate 7-cm radius. From
2
figure 3, we know that the zone 2 action integral is 0.25 x 106 A2 sec. Table 3 indicates that a 1.61 cm cross-
section of graphite/epoxy can sustain this level of lightning discharge without significant structural damage. Using the
equation above, we find that required laminate thickness would be:
d > h 6 cm 2
S "17 (7"cm) 0.07 cm
Such approaches should be used cautiously, however, since ply orientations, material type, protective paints,
curvatures, and the like all have a possihle effect on the results. Any safety-of-flight applications of graphite/epoxy
structure should be verified by test. Non-flight-safety items may iot require tests; past history on comparable
hardware may minimize the risk of not testing. Also, the probabiit. Af lightning attachment ma) be low and the cost of
repair or replacement may be insignificant.
i -:i "
14-5
For example, the test levels shown in figures 3 and 4 are f or worst case testing, generally reserved f or flight-sdfety
items (ref. 5). Commercial aircraft are struck an estimated once per plane per year. This is a coarse fleet-wide
avwrage. As shown in table 4 (from ref. 6) the probability of occurrence decreases with threat level. Also, some
surfaces are hit more frequently than others; for example, an aileron trailing edge is hit less often than a wingtip. All
these factors should be weighted when considering protection for nonsafety-of-flight graphite/epoxy applications.
If lightning protection is deemed necessary, there are a number of possible protection systems. During the reference 7
investigation the following systems were evaluated:
Each of the graphite/epoxy tests specimens was painted %,th polyurethane after the protection system was applied.
Besides lightning testing for all lightning zones, the inves Ligation conducted the following tests on the test sr ecimens:
0
I
Environmental exposure
e
14UOF at 100% relative humidity for 1,000 continuous hr
Ground-air-ground flight simulation for 1,000 cycles
* Temperature = -65 0 F to 250F (-53.89"C to 121.11 C)
* Humidity to 100%
0 Pressure - sea level to 40,000 ft (12.19 kin)
e Cycle time =1 hr
1
.4. *
*
Fatigue
*
+9,000 to +20,000 lb (4,082.4 to 5,072 kg) !or 1,000,000 cycles (typical)
Fatigue rate 1,800 cycles/min
CONCLUSIONS
Lightn~ng damage to graphite/epoxy aircraft structure is not as extensive as once feared. The electrical cor.ductivity
qualities of graphite, although less than that of aluminum, normally are adequate to conduct lightning curre its except
in the immediate arca of lightning arc attachment. A formula is given to help estimate the degree of struc.ural
damage at this arc attachment point versus threat level and laminate thickness. If safety of flight is involved,
estimates of graphite/epoxy structural damage should be verified by testing.
If lightning protection is deemed n=cessary, a number of conductive protective systems are available and in the
literature.
14-6
REFERENCES
I. Fisher, F.A., and Plumer, J.A., "Lightning Protection of Aircraft," October 1977, NASA Reference Publication
1008.
3. 1975 Conference on Lightning and Static Electricity, at Culhum Laboratory, England, April 14 to 17, 1975.
5. "Lightning Test Waveforms and Techniques for Aerospace Vehicles and Hardware," Report of SAE Committee
AE4L, June 20, 1978.
6. Cianos, N., and Pierce, E.T., Stanford Research Institute, Menlo Park, California, "A Ground-Lightning
Environment for Engineering IJsage."
7. Hendricks, C.L., Schneider, S.D., and Shumaker, G., "Lightning Protection Techniques for Graphite/Epoxy Aircraft
Structures," Third National Symposium, Society for tne Advancement of Materials and Process Engineering,
Anaheim, California, May I to 4, 1978.
Ii e
iiI
'1
" ! ot
S~15.1
SUMMARY
Carbon ard graphite fibers are kncwn to be electrically conductive. That property
has resulted in damage to electrical equipment from the inadvertent release of virgin fi-
bers into the atmosphere. The rapidly accelerating use of carbon fibers as the reinforce-
ment in filamentary composite materials brought up the possibility of accidental release
of carbon fibers from the burning of crashed commercial airliners with carbon composite
parts. Such release could conceivably cause widespread damage to electrical and electron-
ic equipment. This paper presents the experimental and analytical results of a compre-
hensive investigation by the National Aeronautics and Space Administration of the various
I elements necessary to assess the extent of such potential damage in terms of annual ex-
pected costs and maximum losses at low probabilities of occurrence. A review of a NASA
materials research program to provide alternate or modified composite materials to over-
come any electrical hazards from the use of carbon composites in aircraft structures is
S~described.
INTRODUCTION
The National Aerona,,'-,.s and Space Administration (NASA) has been deeply committed
to the research and developaent of advanced composites for aerospace applications for more
than a dozen years. During the initial years of that research, the NASA followed an ev-
olutionary, rather than revolutionary, approach toward the use of composites on civilian
aircraft. Concurrent with several technology programs, NASA-sponsored flight service
programs involved the selective reinforcement of conventional metallic structures (1).
The modification of the tail cone of a CH-54 helicopter with stringers reinforced with
strips of unidirectional boron/epoxy composites served to increase the vertical bending
stiffness of tne tail cone, while the addition of similar boron/epoxy reinforced stringers
and wing planks to the center wing boxes of two C-130 transport airplanes served to re-
duce the stress levels and increase the fatigue life of the wing boxes. Both modifications
were achieved at significant weight savings compared to the comparable metallic fixes.
Following an Air Force/NASA Long Range Planning. Study for Composites (RECAST) in 1972
(2), several flight service programs which met the principal objectives of RECAST, to build
confidence in the use of composites in aircraft and to point the way to lower costs, were
begun. Several Kevlar/epoxy fairings were chosen for flight service evaluation on a num-
ber of commercial L-1011 transport aircraft, while aft pylon skins of boron fiber and
were put into
aluminum Twn service on three DC-10 aircraft to replace production titanium
skins. flight service programs involving secondary aircraft parts built of carbon/
service capability of com-
posite comp.;ites
epoxy parts. have successfully
Spoilers with carbon/epoxy skins the
demonstrated flownterm
have long a cumulative total of more
than 1 million hours in active service on Boeing 737 airliners. DC-10 upper aft rudders
built principally of carbon/epoxy materials have also been flown without any major prob-
lems in commercial airline serviue.
9k In 1975, NASA began a program to accelerate the development of several technologies
4 for improving substantially the fuel efficiency of commercial air transports. A vital
part of that Aircraft Energy Efficiency (ACEE) effort (3) is the composite structures de-
velopment program. That composite program involves the development, under NASA contracts,
of six aircraft components (Figure 1) by three major transport manufacturers, Boeing,
Douglas, and Lockheed. Carbon fibers are the principal, though not sole, reinforcement
material to be used in those composites. This follows the emergence from the early tech-
nology add flight service programs of carbon (also referred to in this paper as "graphite")
fibers as the filamentary reinforcing material of choice for aircraft composite parts by
reason of a combination of desired properties together with low cost, high production,
and weight savings potentials.
NASA's accelerated efforts to promote the use of composites as a way to fuel-efficient
commercial air transports have been matched not only by the progress achieved in utiliza-
tion of carbon composites in new U. S. military aircraft, but also by independent programs
by major U. S. airframe manufacturers designed to use advanced coi.kosites to save weight
in their present and future airplanes (4). The concomitant emergence of carbon fibers
as a useful material for other industrial production uses, such as in sporting goods man-
ufacture, has led to the realization that this is a material destined for multi-million
"kilogram quantities within the next decade (Figure 2).
A possible barrier to the optimistic future for carbon fibers appeared with the re-
port (5) that there was an uncertain risk associated with the widespread use of carbon
3 1fibers due to their good electrical conductivity. Since they are extremely fine (about
3
8 microns in diameter) and lightweight (about 1.7 grams per cm ), they can be transported
by the wind for great distances. In contact with electrical devices, they could create
15-2
a number of adverse electricail effects, includinq resistive loading, short circuits, and
arcing, which could lead to electrical outages or destruction (Table I) (6). Several ac-
tual such incidents gave credence to the potential for damage. Several crashes involving
U. S. military aircraft with composite parts built of nonconductive boron fibers demon-
strated (Figure 3) the likelihood that similar crash and subsequent fire events involving
aircraft with carbon fiber composites might release free carbon fibers into the atmosphere
after the restraining influence of the resin matrix was removed from the composites
through oxidation by the fire. Furthermore, carbon fiber manufacturers and processors
have been aware for some time of the electrical problems which could occur from the in-
advertent release of carbon fibers during industrial operations, though they soon learned
to include simple, yet necessary, preventative steps in their normal manufacturing pro-
cedures (7).
The United States Government has an extensive interest in the widespread use of car-
bon/graphite composite materials and any potential hazards resulting from their use. A
number of recent programs have accelerated that interest. Some examples are: NASA's ACEE
program to expand the use of composites in civil aviation, federal procurement require-
ments for lightweight composite materials in high performance military aircraft, and fej-
eral automotive fuel consumption standards which require weight reductions that may be
met only by the significant utilization of composites. Consequently, the importance of
those programs contributed to a decision to conduct a government-wide program to study
and quantify any risks which wou:d result from the widespread usage of carbon fibers.
J. The United States Federal Act-on Plan to deal with potential carbon fiber problems
, begun in early 1978 (8). The overall program, which was ccordinated by the Office of
-. -nce and Technology Policy (OSTP), involved about ten government departments and agen-
c,'&shown n Figure 4. Although NASA was to assist other departments in several ways,
.t. main responsibility was to conduct a thorough assessment of the risks of electrical
and electronic damage t~iat might result by the inadvertent release of conductive carbon
fibers from crashes of civilian aircraft with carbon composite parts. That responsibility
also carried an obligation to assess the need for protaction of civil aircraft from free
carbon fiber. An additional charter for NASA was to take part in, and coordinate, a pro-
gram of research and development of modified or alternate composite materials which would
be harmless from the standpoint of electrical hazards, yet which would satisfy the require-
ments of advanced composites for aircraft structures. While the R and D program for al-
ternative materials was assigned to all of NASA's research centers, the task of assessing
the possible hazards from the use of 'carbon fibers was assigned to NASA's Langley Research
I Center.
A Graphite Fiber Risk Analysis Program Office was formed at Langley to conduct the
risk analysib. The scenario that was adopted as being of interest for the accidental re-
lease of carbon or graphite fibers from civil aircraft is illustrated in Figure 5. This
scenario was hypothesized to encompass the consequences of accidents involving civil trans-
port aircraft, usually near major airports, where the efflux from the burning of composite-
carrying aircraft has the opportunity to be distributed downwind and thus to contaminate
not only the nearby airport environs but also numerous distant facilities as pictured in
the illustration. Another dimension of the risk analysis is afforded by the risk analy-
sis flow chart, shown in Figure 6, which includes all of the elements of importance which
can lead to damages from the release of conductive carbon fibers. Inasmuch as NASA's
risk analysis was organized and conducted according to the elements in the flow chart,
this paper will be presented in the same manner.
! DISCUSSION OF RESULTS
Sources of Released Carbon Fibers
*., The Graphite Fiber Risk Analysis scenario assi'rmed carbon fibers to originate princi-
pally from fires from crashed commercial aircraft with carbon composite parts, as repre-
sented in Figure 7. However, non-crash events such as in-flight or ramp fires were also
;J considered. At the start of this program, there had been no known crashes of commercial
q aircraft with any carbon composites in their structures. In the absence of actual ex-
A perience, an investigation of ways to simulate the conditions of commercial aircraft
crash fires was deemed essential in order to predict the quantities and forms of carbon
fiber residues which might be released.
: The earliest experiments in release of carbon fibers from burned composites classi-
fied the principal types of fiber residues as shown in Figure 8 (9). Single fibers had
the potential for the greatest range of distribution due to their very slow settling rate,
Swhile he clustering together of a number of single fibers gave a form of efflux which
would fall much faster than individual fibers. A third class of residue consisted of yet
4faster settling strips of fibers, generally resulting from a single ply of a crossplied
composite, with the fibers being bound together either by incompletely burned resin or
he char formed by the burned resin. A fourth broad class of fibrous residue was frag-
ients of the composite, widely varied in size and shape. These fragments, which were
so dense that they were rarely found beyond the immediate vicinity of the fire, generally
resulted only from a substantial impact to the burning or burned composite. Rarely were
they formed as the result of a simple fire.
The emphasis of this carbon fiber source study was placed on the generation of single
carbon fibers from fire tests. First of all, the spread of single fibers was considered
to be the most extensive due to their buoyancy, and damage to electrical and electronic
equipment spread over a broad area of population was felt to pose the greatest possible
15-3
economic loss. And secondly, single fibers were considered to be the form of residue
which would be most capable of penetrating the filters and cases of equipment to reach
the vital interiors. Accordingly, most of the fiber release data was gathered and uti-
lized as the percentage of single fibers released (over one millimeter in length) based
on the mass of carbon fiber initially present in the composite test specimens. The ini-
tial experiments which defined the broad classes of residues involved the destruction of
the burned composites with explosives (10). Obviously, such a procedure lacked credibil-
ity for representing the typical commercial aircraft crashes. Furthermore, although the
fire plus explosive test showed substantial amounts of the highly buoyant single fibers
being released, there was an uncertainty as to both the amounts and the forms of residues
which might be generated from the simulated burning of composite aircraft parts without
the explosion step.
The testing program was conducted at six locations on contract to NASA-Langley. The
majority of the small scale composite burn tests were carried out at the U. S. Navy's
Dahlgren (VA) fire test facility, which was a completely enclosed environmental chamber,
approximately 100 square meters in area, in which small samples (up to 0.1 square meter
in size) were burned with a propane gas burner and a variety of disturbing effects were
applied to the fibrous residue after consumption of the matrix resin. The test samples
were generally either flat composite plates or smalt specimens cut from prototype compos-
ite aircraft structural components. The principal advantage of this test facility was
the complete containment of all the fibers released, which were collected by allowing
them to settle onto highly adhesive deposition papers laid out on the floor of the cham-
ber. The fiber output from the tests was then counted, generally by optical microscopic
methods, in order to analyze the test results. Additional small scale test support was
provided by the AVCO Corporation's fire test facility in Lowell, MA (11) and a NASA-Ames/
Scientific Services, Inc. test facility at Redwood City, California (12).
In addition to the small-scale laboratory test, valuable fiber release information
was acquired from outdoor composite burn tests conducted by TRW Inc., Redondo Beach,
California under U. S. Air Force sponsorship (13), for the purpose of studying not only
the nature of fiber release from burning composites, but also to verify predicted dissemi
nation patterns over broad areas. Confirmation if small-scale test data was also obtained
from large-scale demonstration testing conducted in the U. S. Navy's Dahlgren, VA tube
test facility and at the U. S. Army's Dugway (UT) Proving Ground.
Laboratory testing addressed the effects of the following variables on the amounts
and characteristics of carbon fibers from burned composites:
Type of fire fuel (jet fuel, propane, natural gas)
Nature of fire (fuel-rich, fuel-poor)
Duration of fire
Disturbances to residue during and after fire
Composite thickness and configuration (crossply, woven, unidirectional, etc.)
Composite surface and edge effects
Types of composite materials (fibers and resins)
Composite quality
fibers the
3. Despite
of carbon from oft-quoted
composite aircraft parts were of
"indestructibility" graphite
consumed in representative fuelmasses
fiber, substantial fires
considered to be typical of jet aircraft crash fires. This reflects the fact that the
"carbon/graphite fibers in use today are not graphite to any great extent. However, fu-
ture use of the mora graphitic, high modulus type would undoubtedly lead to less consump-
tion of fibers in fires. The mass loss which was observed was manifested by a reduction
"in the thickness of the fibers, but there was probably some complete consumption of fibers
when released in the hotter regions of the fire as well as from the burning composite
itself.
154
N The fiber materials released from a fire form a cloud which moves with the velocity
of the wind and in the same direction as the wind. At a short distance from the fire,
the effects of diffusion create Gaussian distributions in the concentrations of fibers
along the direction of travel and across the spread of the cloud. The rate of spreading
and also the maximum altitude of the fibers within the cloud are determined by the weath-
2 er conditions. The ground level exposure can be described by a series of contours or
"footprincs" which link points of equal exposures to either single fibers or lint (clus-
ters of fiber). As Figure 12 shows, overcast or nighttime conditions generally produce
longer, narrower contours, with fhll-out distances of up to 100 kilometers. Fall velocity
has a direct effect on distance, with the heavier lint falling cut in proportionately
shorter distances. Conditions typical of sunny weather tend to give shorter but broader
contours.
The single fiber coverage has been plotted in Figure 13 for a number of quantities
of released carbon fibers and for a number of appropriate exposure levels. This exposure
andlysis allows the rapid determination of just how large a geographical region would be
exposed to a certain exposure level of carbon fibers, as the result of the release of a
specific weight of single carbon fibers. For example, if 40 kilograms of single carbon
. fibers with an exponential distribution and a mean length of 2 millimeters were released,
"I an area about the size of the suburb of a city (about 107 square meters) would be exposed
to 106 fiber-seconds per cubic meter. The fiber analyses represented in Figure 13 were
developed from dispersion models. They showed that a certain quantity of released carbon
fibers would be dispersed over a much broader area than had been estimated initially,
but the fiber concentrations would be proportionately lower. The quantity of fiber re-
leased in the worst case prediction of this risk analysis is shown by the dotted line.
I- Fiber exposure levels ranging from 105 to 108 fiber-seconds per cubic meter can be ex-
pected to result in some failures of electrical equipment. A more detailed discussion
of equipment failures will be given in a later section on Vulnerability.
_Sa
i. 15-5
properly mounted: even a small gap between the filter and its mounting can redtice the
ilter efficiency by an order of magnitude or more. Air velocity can also have a sub-
stantial effect on the number of fibers passing a filter, with a ten-fold increase in
air velocity resulting in a similar increase in passage of fibers. Air filters (or water
separators which serve the same purpose) in transport aircraft have extremely small
transfer functions (02x 10- ). However, as much as three-quarters of the fiber which
approaches an open aircraft avionics bay door could pass into that compartment while the
plane is on the ground. Entry of carbon fibers could be prevented, in the event of a
nearby incident, simply by closing the door. Nevertheless, filtration and natural bar-
riers to the transmission of carbon fibers were shown to provide alleviating effects to
the electrical hazards of the fibers.
Vulnerability
The investigaticn of the elements dealing with the vulnerability, or susceptibility,
of electriual and'electronic equipment to damage from carbon/graphite fibers was preceded,
perhaps, by the most speculation about the seriousness of the problem. While several in-
cidents of damage caused by virgin carbon fibers in or near carbon fiber manufacturing
operations were a matter of record at the onset of NASA's study, there was apparently
no known instance in which fibers previously incorporated into a composite had been re-
leared in an accidental fire and had subsequently caused adverse electrical effects.
Consequently, a number of unknowns regarding the non-virgin carbon fibers had to be
studied and their effects on damage potential analyzed.
The types of effects that were expected from exposure of equipment to fibers were
listed previously in Table I. The categorization shown is very gross, but serves to
point out those general regions which were of the most concern. In the low voltage and
low power region, a carbon fiber has the capability to maintain a high resistance short
without burning out. In that event, the equipment could malfunction although the com-
ponents incur no local damage. Tne region of medium voltage and high power causes some
concern since the fiber acts aL a trigger to potential arcing, and the arcing may be
sustained. The result can be blown fuses, stressed components and flashovers at insula-
tors. The third region of most concern is that of both high power and high voltage,
where carbon fibers (as is the case for many foreign materials) can lead to corona, and
initiate arcing severe enough to disrupt and to damage equipment catastrophically.
The electronic equipment failure model used to describe the vulnerability of elec-
tronic equipment to carbon fibers has been shown in Figure 17, where the probability (P)
of failure is: P = l-exp [-E/9] , where E is the mean exposure required to fail the
equipment during testing and E is the exposure endured by the equipment. At the E expo-
sure level, the probability of failure of a piece of equipment is 0.632, while at an ex-
posure of one-tenth the E, the probability is only .095, meaning about 9 cr 10 pieces of
equipment of the same type would fail out of 100 exposed.
The NASA Vulnerability Test Program relied principally on two test methods to assess
the susceptibility of electrical and electronic equipment to carbon fibers. The simplest
technique involved the use of a fiber simulator. This simple device was developed to
simulate the electrical characteristics of a carbon fiber. If the electronics or the
equipment to be tested had a limited number of nodes, 50 or perhaps no more than 100,
it was possible to probe the equipment and so be able to tell if a carbon fiber (as rep-
resented by the fiber simulator) could cause the equipment to malfunction or fail. Rep-
resentative electric appliances which were tested in that manner have been listed in
Table II (18).
The most direct method to test electrical equipment for vulnerability to carbon fi-
bers involved direct exposure of the equipment in a controlled chamber. An example of
the several chambers used in this program is the one located at NASA's Langley Research
,= Center, pictured in Figure 18. The operations scheme involved placement of the test
article in the chamber, while a known quantity (monitored by fiber sensors) of carbon fi-
bers of a desired length were chopped from continuous fiber tow and aspirated into the
exposure chamber. The fibers were then kept in suspension until the test article mal-
functioned or failed, or until a maximum exposure level (usually 108 fiber-seconds per
cubic meter) was reached without failure. A number of replicate tests were conducted
for each test item, whether it failed or not. A thorough cleaning of the equipment,
such as by removal of fibers with a vacuum cleaner, was conducted between tests. A sum-
mary of the articles tested in the NASA or other chambers is presented in Table III (19).
Over 150 individual articles were tested for electrical/electronic susceptibility
to carbon fibers by the fiber simulator and test chamber methods. The results in Tables
II and III indicate that many pieces of equipment were invulnerable to carbon fibers.
Most household appliances with 110-volt circuits proved to be unaffected. Enough pieces
of consumer equipment were selected to be representative of about 75% of the market value
of consumer gcods. Very little 220-volt testing was done, but it was expected that 220-
volt single-phase circuitry would be about as resistant to carbon fibers as 110-volt,
I . since 440-volt single phase equipment was also proven to be relatively unaffected. (How-
ever, arcs could be sustained for 440-volt, 60 hertz industrial power using three phase
transformer supplies capable of dclivering currents in the range of 400 to 1500 amps,
when arcs were i.nitiated. An idea of the exposures leading to failure of some equipment
can b seen from Figure 19 (19). Most equipment was vulnesable at 10 -i0 fiber-seconds
per ms for longer fibers, and near 100 fiber-seconds per m for very short ones. No
equipment failed in the lower left hand region of the figure outlined by shading.
.i4 -
15-7
Figure 20 (19) illustrates an important factor influencing the effect of carbon fi-
bers on electrical circuits. The tests were conducted with three different fiber lengths:
short (N3 mm), medium (,v7 mm), and long ('x12 mm). It is apparent that the longer fibers
were much more effective in causing malfunctions than short ones. As was pointed out in
the section on Source, most fire-generated carbon fibers are very short, with mean lengths
usually between two and three millimeters. Therefore, the exposure levels causing equip-
ment failure are quite high for the fibers released in real-life situations.
Another important relationship which was established by fiber chamber tests was
that associated with the resistance of the carbon fibers themselves. Figure 21 (19)
shows the effect that the fiber resistance had upon the critical exposure levels for
three pieces of equipment. The stereo amplifier was an order of magnitude in exposure
less vulnerable to DE 114, a high resistance, low temperature-processed carbon fiber,
than it was to the T300 fiber in common use today. Similarly, a color television set
and an air traffic control transponder were from 1-1/2 to 2 orders of magnitude of ex-
posure less vulnerable to failure from T300 fiber than they were to two highly conduc-
tive, very high modulus fibers GY70 and HMS. Studies also concluded that fibers re-
leased from composites by fire had resistivities unchanged from virgin fibers and the
damage potential of such fibers was the same as for the raw fibers. Another concern was
that of post-exposure vulnerability. Most of the testing of equipment in Table III was
done in the "on" condition. There was some concern about whether or not equipment which
was exposed while "off" would fail subsequent to turning it "on". A test scheme irvolv-
ing 200 hours of testing of a color television set and a stereo amplifier, including many
"on-off" cycles, indicated that post-exposure failure was not a significant problem.
hazards. Under NASA sponsorship, the National Bureau of Standards examined a large num-
:1 ber of household equipment items for susceptibility to failure and/or shock hazard. As
mentioned before, 110-volt household appliances were generally invulnerable to carbon
fibers. However, at extreme exposure levels some appliances were susceptible to carbon-
fiber induced shorts to the external appliance case where potential shock hazard can
exist. The most susceptible equipment for the shock potential was found to be the com-
.1 !!'iaccident mon household
rates toaster
in 1993 (21). An analysis
(the year for whichbased on the
the risk analysis was carbon
projected fiber indicated
performed), usage and
1 less than one potential shock hazard per year would be caused by accidental carbon fiber
release. Furthermore, it was predicted that the short current would not be lethal since
the fiber would burn out (that is, using the 30 million psi modulus fibers in use in
1980).
Demonstration Testing
A series of tests were conducted (22) in a large, tubular fire facility to demon-
4 I strate an agreement between the susceptibility of electronic equipment to carbon fibers
generated from burning composites in a jet fuel fi-e and the vulnerability of the same
equipment to clean, virgin fibers in the Langley exposure chamber. The unique fire
chamber was a modification of a portion of a long, shock tube located at the Naval Sur-
face Weapons Center, Dahlgren, Virginia. A photograph of the 750-meter long tube is
pictured in Figure 22. A 275-meter section of the tube was utilized, with a 1.22-meter
square commercial jet A fuel fire being burned at a location where the tube was 4.6
meters in diameter. Composite specimens were burned in the fire and the fire plume was
pulled through the last 275 meters of the tube by up to six large fans. A water fog
spraying down from the top of the tube served to remove carbon fibers from the smoke
fw
1. plume, which exited from the 7.3-meter (24-foot) diameter end of the tube. During the
equipment exposure tests, six identical fan-cooled, unfiltered stereo amplifiers were
situated on a target table at 220 meters (700 feet) from the fire. The amplifiers were
,,zV---'--
15-8
in an operating mode during the fiber release fire test. Strips of carbon fiber/epoxy
composites were placed in a wire mesh basket, which was rotated in the middle of the fire
during the entire period of the test. The actual failure of the amplifiers have been
represented in Figure 23 by the step-wise solid line plot. The first four amplifiers
had failed after the first 600 seconds of the tgst. Those failures have been indicated
by the step-up at an exposure of about 6.5 x 1V fiber-seconds per cubic meter, with a
fifth failure occurring when an exposure of 2.4 x 10 fiber-seconds per meter 3 had been
reached, and the final amplifier failed at a level of about 3.3 x l0b fiber-seconds per
meter-. The experimental failures in the shock tube test matched very well the failures
predicted from the Langley chamber-derived exponential probability curve superimposed on
the figure.
A series of large scale outdoor demonstration tests was conducted at the U. S. Army's
Dugway Proving Ground in Utah (23). The series consisted of two types of tests: source
tests designed to measure the extent of fiber release from burning carbon/epoxy compos-
ites in a large JP-4 fuel fire and plume tests which were intended to not only capture
carbon fibers and so determine the amounts released, but also to disseminate the released
fibers over an area so large as to realistically simulate the dispersion expected from
the crash and burning of a commercial air transport with carbon composite parts. The
tests were carried out using a 10.7-meter diameter fuel pool size and 11.4 cubic meters
of JP-4 aviation fuel. Duration of the fires was nominally 1200 seconds. About 45 ki-
logram quantities of real and test aircraft parts of cation/epoxy composites were placed
on an elevated steel mesh table above the fire pool.
The source tests were conducted during periods of very low wind speeds (less than
0.4 meters per second) in order to allow the fire plume to rise vertically from the fire.
A large number of steel mesh sampling devices which captured released carbon fibers on
the mesh screen within a cannister were suspended in an array above the fire. The sam-
plers were suspended from cables rigged from four 60-meter high towers, arranged in a
65-meter square around the fire pool.
The dissemination tests were conducted in the same manner except that wind speeds
from 2.7 to 5.4 meters per second were desired, and a wind direction of 3200 1 350 was
requirei. This permitted the fire plume to pass throunh a huge "jacobs ladder" (Figure
24). (The 169-meter high Washington Monument has bee. shown to scope the size of the
undertaking). This "jacobs ladder" vas constructed fEom 2.54 mm Kevlar
rope with
horizontal and vertical spacings of 15.25 meters. The 305-meter by 305-meter network
was suspended from a catenary whi6h was lofted by two U. S. Air Force 1270 m3 balloons,
with stabilizing tether lines placed out in all directions. The net was placed 153
meters from the fire. Manf sampling devices of several types were mounted on the net.
These included flat plastic rectangular frames, with their 0.29 m x 0.23 m openings
covered with 1 mm mesh fabric coated with a sticky substance to cause fibers to adhere
to the mesh when the samplers were placed at the intersections of the rope, normal to
the flow of the smoke plume. Other samplers!included mesh filters in cardboard cannis-
ters similar to the steel samplers suspended over the fire, eight high voltage electri-
fied grids instrumented to discharge when contacted by fibers, open-ended 0.085 meter
diameter cans with adhesive-coated fabric mesh spread over one of the open ends, fiber
collection pumps and filters used to monitor the air for excessive concentrations of
respirable-sized fibers, and light emitting diode detection devices. As the fire plume
passed through the suspended "jacob's ladder", the fibers were detected or collected by
the array of monitoring apparatus which was then analyzed subsequent to the test. In
addition, the dissemination of fibers was monitored by means of both deposition sticky
papers and vertically mounted open-ended mesh can samplers spread out at appropriate in-
tervals for distances of up to 19 kilometers from the fire in the direction of the wind
ii
flow.
A summary of the results of both the source and the dissemination tests has boen
presented in Table IV. Variance from the laboratory tests reported in the Source sec-
i tion of this report was noted only for the average fiber lengths. The average lengths
of 5.0, 4.4 and 5.2 millimeters from the three dissemination tests were somewhat higher
than the lengths obtained for the laboratory tests. However, the average length (3.2
mm) from the source test was in keeping with the laboratory test findings. The average
fiber diameters of 4.1 to 4.7 micrometers indicated a substantial oxidation of the fi-
bers from their normal 6 to 8 micrometers in the virgin state. The weight percent of
fiber release was in concert with many of the laboratory results.
11 Conclusions pertinent to the risk analysis, based on the results of the large scale
demonstration tests at Dahlgren and Dugway Proving Ground, are summarized below.
A
maximum of 0.5% of carbon fiber, based on the amount initially present in the
composite specimens exposed to the fire, was released in the best Dahlgren shock tube
fire and equipment exposure test. However, since that was a long duration fire (over
S12,000 seconds) and the fiber release was forced, the maximum of 0.19% released from
the Dugway demonstration tests was considered more representative of predicted fiber re-
lease from commercial air transport fires. Therefore, the figure of 1% carbon fiber re-
lease used in the risk calculations was quite conservative. The mean fiber length (2
mm) from the Dahlgren test was in close agreement with the value used in the risk analy-
sis, but some of the mean lengths from the Dugway large scale tests were somewhat longer
"than the mean length used for the risk analysis. And finally, the Dahlgren demonstration
[ Registered trademark for DuPont's Aramid fiber
15-9
tests established that the vulnerability of equipment to fire-released fibers agrees with
the vulnerability of the equipment to virgin, unburned carbon fibers, thus justifying
the use of fiber chamber test data in the risk calculations.
Facility Surveys
Surveys of 62 public, utility, commercial and industrial installations in the United
States were conducted (24) in order to develop a sound foundation for the use of census
data generated in the analysis of the overall risk to the community from the use of car-
bon fibers in civilian aviation. A summary of the number and types of installations of
the four major classes has been summarized in Table V. Emphasis was placed on three
main elements:
(a) Determination of data for use in modeling the economic impact of fiber-induced
failures;
(b) Identification of the sensitivity of life-critical or emergency services to the
fiber hazard; and,
o Life critical functions could be excluded from any impact on the risk analysis.
o Emergency services would suffer no interruption. Any economic impact would be
limited to specific items of equipment.
o Utilities would suffer no system loss. The economic impact would be confined to
local outages and repairs.
o Commercial institutions, such as banks, stores, etc., would incir no interruptions
r~l to critical
ripheral operations.
equipment. Any adverse impact from carbon fibers would bu limited to pe-
i The primary objective of the risk analysis was to estimate the risk to the nation
over the next 15 years (from 3978) resulting from the use of carbon composites in civil
aircraft. A secondary purpose was to provide a framework for decision making on com-
posite material usage, material modification, and protection schemes. Two contractors,
ORI and Arthur D. Little, Inc. were selected to develop independently methods to numer-
ically evaluate the potential losses due to failures of electrical equipment from air-
borne carbon fiber contamination originating from civilian aircraft crashes. The risk
computations were conducted in two phases. ORI developed a risk model based on the 9
largest hub airports in Phase I (25) and proceeded to translate the risk profiles for a
number of individual airports into a national model in Phase II (26). Arthur D. Little,
' Inc. developed a preliminary national profile from 26 major airports in the first phase
(27) and followed up with a number of refinements to the national risk profile in Phase
V 11 (28).
15-10
The available commercial air transport accident records of the National Transpor-
tation Safety Board (NTSB) were augmented by information from the three major U. S. air-
frame manufacturers. They supplied detailed data on accident characteristics, such as
fire duration and severity of damage to aircraft, that would affect carbon fiber release
conditions. Projections on the extent of carbon fiber usage on U. S. commercial air-
craft were also generated by the aircraft manufacturers. Such projections included the
numbers of planes expected to be in service in 1993, by three types: small, medium, and
large jets, as well as the ranges of quantities of carbon fiber composites which were
predicted to be used on those planes. Among the criteria for determining the aircraft
crash scenarios were the operational phases during which the accidents occurred. Figure
25 summarizes NTSB data for 1968-1976 129) indicating that almost half of the severe ac-
cidents accompanied by fire occurred during landing, while a quarter of the accidents
with fire occurred during takeoff. Furthermore, 60% of the accidents happened at the
airport and 80% were within 10 kilometers of airports. Attention was also focused on
26 large hub airports which accounted for nearly 70% of U. S. emplanements. And lastly,
3.8 severe fire accidents were predicted annually in the United States. Based on pre-
dictions that 70% of the jet fleet would be using carbon composite parts by 1993, 2.7
fire accidents per year were projected for 1993 for such aircraft.
Figure 26 (27) shows the sequence of events which were modeled in order to describe
the carbon fiber risk phenomenon. The simplified event tree logic which was followed to
arrive at the local risk profile has been depicted in three sequential figures, Figures
27a, 27b, 27c (30). Random selections are used during many phases of this event tree for
selecting the paths. At various points, the random selection leads to inputs from vir-
ious elements of the entire program. For example, if an accident is randomly selected
(Figure 27b) which involved an explosion in addition to a fire, then 3-1/2% single fiber
release will be used. If the accident chosen involved fire only, then 1% single fiber
release was used in the model of that accident. When the path of the event tree reaches
Figure 27c, the areas of the city and/or countryside affected have been defined. Input
from the elements of transfer function, vulnerability, and facility surveys then permit
the determination of cost impacts from the accident. Examples of costs are repair or
replacement of equipment, downtime, product losses, etc. Using a selocted historical
number of accidents each year (1u3 to 6), the random selection of nodes in the event tree
and the cost calculations are repeated for each accident and the cost is summed to obtain
one estimate of the national cost. One estimate, however, is insufficient to obtain a
statistical distribution of estimates, .. o the national risk calculations must be repeated
a larga number of times.
The results of the annual risk profiles for economic losses due to commercial air
transport fires involving carbon fibers have been given in Figure 28. The Phase I pro-
file was assessed one year earlier than those from Phase II. The markedly lower risks
in Phase II were attriu.Lable to a number of refinements in single fiber release (5%
fire, 25% fire-explosion in Phase I vs. 1% fire, 3-1/2% fire-explosion in Phase II), a
ten-fold decrease in infiltration due to use of experimental transfer function data, and
extreme diminishments in equipment susceptibility due to shorter fiber lengths and
higher mean exposures to failure. An increase in predicted carbon fiber usage in air-
craft manufacturing made a slight positive contribution to the risk profile.
As the risk profiles in Figure 28 show, the expected annual risks to the United
States due to the predicted use of carbon composites on civil aircraft in 1993 are cer-
j tainly less than $1000 per year. The chances of national losses reaching significant
levels are extremely small. For example, both the ORI and Arthur D. Little (ADL) models
indicate (at the crossover for the plots) that an accident resulting in $5000 damage
from carbon fibers would only occur every 40 years! Although a trend toward significant
use of carbon fiber composites on general aviation aircraft has not yet emerged, separate
risk computations dealing with forecasts of up to 55 kilograms of such composite per
plane were made. The conclusion from that study was that it was extremely unlikely
there would be a substantial dollar loss due to carbon fiber releases in general avia-
tion accidents. Further diminishing the concern for carbon fiber hazards is the fact
that loss of life from carbon fiber electrical events is virtually non-existent.
Material Modification
As pointed out in the Introduction, a secondary responsibility assigned to NASA
inder the Federal Action Plan was to investigate alternative or modified composite ma-
A .erials which would lessen or eliminate electrical hazards as a consequence of the use
of composites. This responsibility was assigned to NASA's research centers since new
materials research and development fit into the existing charters of the centers' base
technology programs. All of the installations (Langley, Ames, and Lewis Research Cen-
ters, Marshall Space Flight Center, and the Jet Propulsion Laboratory) had knowledgeable
line organizations with composite materials expertise so that a minimal impact on their
existing research programs resulted.
The general objective of the resulting program was to develop composites which re-
duce carbon fiber electrical risks while retaining or improving the structural proper-
V ties of resin matrix composites which make them desirable for use in aircraft struc-
tures. A NASA workshop on modified and alternate materials was held at Langley Re-
,\ p~search Center (31) in March 1978 as a means of making the composite materials community
aware of the carbon fiber hazards program and to solicit the ideas of goveritment, in-
dustrial and academic representatives. The cooperative efforts of the workshoj. resulted
in a program with the following elements:
2
.- Y
?h
15-I1
Test Methods
Hybrid Composites
Fiber Gasification
Modified Epoxy Resins
Fiber Coatings
Alternate Matrices
High Resistance Fibers
New Fibers
The element of test methods was given the highest priority since it was essential
that investigators be able not only to determine the effectiveness of their new or modi-
fied materials, but also be able to compare them with other modified materials. However,
since research could not be delayed pending the development of standardized test methods,
several di'ferent tests were developed for deterLrining fiber release at several of the
research centers. One rather simple test device was developed at NASA's Ames Research
Center (32) and it was utilized to a limited excent for comparing the relative amounts
of single fibers which were released from burning composites. The device burned rela-
tively small composite specimens, approximately 25-30 cm2 in size, with a heat flux
fairly realistic for a fuel fire, after which the composite was impacted with a pneumat-
ically driven steel ball. The released fibers could then be collected and measured.
The zoncept of hybrid composites to minimize the release of conductive fibers was
considered to be the most promising approach by the workshop participants. The hope that
outer, or alternatin*g, plies of a nonburning fiber reinforcement used together with plies
of carbon fiber would help contain free carbon fibers appeared to be valid in preliminary
tests, especially for composites which were just burned without any severe disturbance.
Although the concept has not been evaluated conclusively, there are indications (33)
that the type of disruption applied to the residual post-burned fibrous mass may be crit-
ical in determining the validity of the hybrid concept.
Fiber gasification aas a novel approach which involved the deposition onto carbon
tibers n cerl.Cn metallic ions which would catalyze the compl..te consumption of the
fibers when exposed to flame. Preliminary results (34) with such contaminants as csl-
cium and bariuz acetates were promising, although much additional rteearch was required
to prove the practicality of this potential solution to the problem.
The modification of ti'e epoxy resins used in composites was attractive from the
standpoint of promising a minnium disruption of current applications and the least re-
qualification of the modified composites. It was early ;4cognized in the determination
of fiber release from carbon-epoxy composites that epoxies are always converted in fires
to a small amount of char which serves to bind individual carbon fibers together, thus
preventing their release for some time. Some promising modified epoxies were uncovered
by changing the chemistry, catalysts, and blending to give mua.j higher char yields (35).
The same thrust toward higher char yield res -, -:isthe basis for much of the attention
paid to another element in the program, that of ?l.einatc matrices.
The general objective of the fiber coatings element of the program was to deposit
coatings onto existing carbon fibers and thus to render them nonconducting. Among sev-
eral coating materials such as silicones, boron nitride, silicates, boron carbide and
silicon carbide (36), (37) the latter offered the most promise. A ten-fold increase in
resistance of carbon fibers was afforded by a 0.1 micrometer c-'ating, but oxidation of
"the silicon carbide to silicon dioxide at 1273 K gave six orders of magnitude increase
in resistance. However, preliminary tests indicated some undesirable effects on the
properties of composites made from the coated fibers.
The last two elements of the materials modification program, nonconductive fibers
and new fibers, were considered to be the most long-term in nature and, thus, were rel-
atively low in priority. When the carbon fiberr were oxidized to carbon oxide fibers,
the resistivities of the resulting fibers were much higher (up to 105 ohm/cm). However,
the degradation of fiber properties was excessive. As a nor;conductive fiber with me-
chanical properties quite similar to carbon, boron nitride fibers were being studied
prior to the emergence of carbon fiber electrical hazards and those early efforts were
revived and augmented as a result of this new program. At least one high modulus, non-
conductive organic fiber was studied as a hoped-for replacement for carbon fibers.
The materials modification program was undertaken with the full knowledge that the
chances of replacing, improving, or even modifying, an existing industrial product srch
as carbon fiber which had undergone years of industrial development were very remote.
It is not possible to claim any positive results from this program at the present time
since the program has been underway for less than two years, and some of the leads are
still being actively pursued. The findings of extremely low risks as a result of the
afore-described carbon fiber risk analysis has certainly diminished the necessity for
new materials research. Nevertheless, the incentive which engendered the modified
materials research program could well prove to have been the driving force instrumental
in the successful development of one or more exciting new composite materials in the
future.
"Concluding Remarks
A comprehensive assessment of the possible damage to electrical equipment caused
ffi by accidental release of varbon fibers from burning civil aircraft with composite parts
has been completed. The study concluded that the amounts of fiber expected to be re-
15-12
leased were lower than initially supposed: conservative quantities of 1% and 3-1/2%
were employed in the risk computations for aircraft c ash fires and crash fires plus
explosions, respectively. Footprints of carbon fibers determined from dispersion models
were found to be much larger in area than originally estimated, but were much lower in
fiber concentrations. Redissemination, as a source for fiber, was shown to be insignif-
icant. The susceptibility of electrical equipment tn carbon fibers was low for current
structural fibers. Consumer appliances, industrial -Lectronics, and aviation instrumen-
tation were relatively invulnerable to carbon fibers. The overall risk costs were shown
to be extremely low: the expected annual cost was lers than $1000 and it was predicted
that there was only one chance in two thousand of exc-eding $150,000 equipment loss in
1993. Furthermore, the potential shock hazard from ctrbon fibers was insignificant, so
risk of life from electrical fiber effects was not considered to be a factor in the
overall risk associated with the widespread use of carbon fiber composites in commercial
aircraft structures.
The results of the NASA risk assessment program axe such that the electrical effects
of carbon fibers should not be considered an impedimenA to further development of carbon
composites in aircraft use. In addition, a program tc :evelop alternate materials spe-
cifically to overcome that perceived hazard is not nece~sa:y.
REFERENCES
1. Brooks, W. A., Jr.; Mathauser, E. E.; and Pride, R. ;N.; Application of Composites
to the Selective Reinforcement of Metallic Aerospace Structures. Conference Re-
print No. 112, Presented at the AGARD Conference on Impact of Composite Materials
on Aerospace Vehicles and Propulsion Systems, Toulovze, France, September 20-22,
1972.
2. Yaffee, M. L.: Use of Composites to Increase. Aviatio.i Week and Space Technology,
September 11, 1972, pp. 87-91.
4. Card, M. F.: Highlights 1979 - Materials. Astronautics and Aeronautics, Vol. 17,
No. 12, 1979, pp. 98-100.
5. A Report of Observed Effects on Electrical Systems of Airborne Carbon/Graphite
Fibers. NASA TM 78652, 1978.
6. Taback, I.: Vulnerability. NASA Conference Publication 2074, 1979, pp. 109-123.
7. McFerrin, J. H.; and Trulson, 0. C.: Electrical Equipment Protection and Waste
Disposal Practices - Carbon Fiber Manufacture. So'ziety of Automotive Engineer
Technical Paper 790034, Congress and Exposition, February 26 - March 2, 1979.
8. Carbon Fiber Study. NASA TM 78718, 1978.
9. Bell, V. L.: Source of Released Carbon Fibers. NASA Conference Publication 2074,
1979, pp. 41-71.
10. Babinsky, T. C.; and Musselman, K. A.: Burn/Blast Tests of Aircraft Structural
Elements. U. S. Naval Surface Weapons Center, NSWC/DL TR 3897, 1978.
11. Alexander, J. G.: Development of a Fire Test Facility for Gvaphite Fiber - Rein-
forced Composites. (AVCO Corporation; NASA Contract NASl-15511). NASA CR 159193,
1980.
i 12. Wilton, C.; Xamburnoff, G.; and Boyes, J.: Fire Testing of NASA Samples - Phase I.
(Scientific Services, Inc.; NASA Contract NAS2-9945) NASA CR 152339, 1979.
t 13. Lieberman, P., Chovitt, A. R.; Sussholz, B.; and Korman, H. F.: Data Reduction and
jA analysis of Graphite Fiber Release Experiments. (TRW Defense and Space Systems
Group; NASA Contract NAS1-15465). NASA CR 159032, 1979.
14. Dumbould, R. K.; and Bjorklund, J. R.: NASA/MSFC Multilayer Diffusion Models and
Computer Programs--Version 5. (H. E. Cramer Co., Inc.) NASA CR 2631, 1975.
15. Turner, D. B.: Workbook of Atmospheric Dispersion Estimates. Rev. 1970 Environmental
Protection Agency, Office of Air Programs Publication AP-26.
16. Trethewey, J. D.; and ffIing, J. H.: Surveillance Sampling of Carbon Fiber Material,
June 1975 - August 197S. U. S. Army Dugway Proving Ground, DPG-FR-79-304, 1978.
17. Paszek, J. J.; Davis, D. I). and Patrick, J. H.: Carbon Fiber Transfer Functions
Through Filters and Enclc'res. U. S. Army Ballistic Research Laboratory
"Report ARBRL-MR-02946, 1980.
18. Lovett, C. D.; and Wise. R. A.: A Study of the Effects of Carbon Fibers on Home
Appliances. National Bureau of Standards Report NBS-IR-79-1952, 1979.
A4
J!.-
15-13
1.9. Taback, I.: Evaluation of Equipment Vulnerability and Potential Shock Hazards.
NASA Conference Publication 2119, 1980, pp. 73-99.
20. Meyers, J. A.: The Vulnerability of Commercial Aircraft Avionics to Carbon Fibers.
(The Bionetics Corporation; NASA Contract NASI-15238). NASA CR 159213, 1980.
21. Meyers, J. A.: The Poteptial of Carbon Fiber Induced Hazards in Household Toasters.
(The Bionetics Corporation, NASA Contract NAS1-15238). NASA CR 159147, 1979.
22. Musselman, K. R.; Pride, R. A.; and McHatton, A. D.: Eleatronic Equipment Vulner-
ability to Fire-Released Fiber Exposures. NASA TM 80219, 1980.
23. Pride, R. A.: Large-Scale Tests. NASA TM 80218, 1980.
24. Butterfield, A. J.: Survef of Industrial, Business and Public Facilities to Estab-
lish Fiber - Induced Failure Consequence. NASA Conference Publication 2119, 1980,
pp 137-149.
25. Pocinki, L. S.; Kaplan, L. D.; Cornell, X. E.; and Greenstone, R: Advanced Risk
Assessment of the Effects of Graphite Fibers on Electronic and Electrical Equip-
ment. Final Phase I Report (ORI Inc., NASA Contract NASI-15379). NASA CR 159027,
1979.
26. Pocinki, L. S.; Kapla-, L. D.; and Cornell, M. E.: Advanced Risk Assessment of the
Effects of Graphite Fibers on Electronic and Electric Equipment. Final Phase II
Report (ORI Inc., NASA Contract NASI-15379). NASA CR 159210, 1980.
27. Kalelkar, A. S.; Fiksel, J.; Raj, Phani P. K.; and Rosenfield, D. B.: An Assessment
of the Risk Presented by the Use of Carbon Fiber Composites in Commercial Aviation.
(Arthur D. Little, Inc., NASA Contract WASI-15380). NASA CR 158989, 1979.
28. Fiksel, J.; Kalelkar, A. S.; and Rosenfield, D. B.: An Assessment of the Risk
Arising From Electrical Effects Associated with Carbon Fibers Released From Com-
merfial Aircraf t Fires.
R15205,1980. (Arthur D. Little, Inc., NASA Contract NASI-15380). NASA
29. The National Transportation Safr.ty Board Annual Review of Accident Data for U. S.
Air Carrier Operations, 196S - 1976. Department of Transportation.
30. Credeur, K. R.: Risk Methodology Overview. NASA Conference Publication 2074, 1979,
pp. 159-171.
31. Dicus, D. L., Compiler: Modified Composite Materials Workshop. NASA TM 78761, 1978.
32. Gilwee, W. J., Jr.; and Fish, R. H.: A Small-Scale Test for Fiber Release From
Carbon Composites. Presented at Society of Plastics Engineers Conference on
"Advanced Composites - Special Topics", El Segundo, CA, December 4-6, 1979.
33. Tompkins, S. S.; and Brewer, W. D.: Preliminary Burn and Impact Tests of Hybrid
~ Polymeric Composites. NASA TM 78762, 1978.
'I 34. Ramohalli, K.: Novel Approaches for Alleviation of Electrical Hazards of Graphite-
Fiber Composites. Jet Propulsion Laboratory Publication 79 - 6 3 , 1079.
35. Serafini, T. R.; Delvigs, P.; and Vannucci, R. D.: High Char Imide - Modified
Epoxy Matrix Resins. NASA TM 79226, 1979.
36. Suplinkas, R. J.; and Ilenze, T. W.: A Study of the Deposition of Carbide Coatings
on Graphite Fibers. (AVCO Corporation, NASA Contract NAS1-15374). NASA CR
159133, 1979.
37. Galasso, F. S.; Veltri, R. D.; and Scola, D. A.: Study of High Resistance Inorganic
Coatings on Graphite Fibers. (United Technologies, NASA Contract NASI-14346).
NASA CR 159078, 1979.
ACKNOWLEDGMENTS
This paper summarizes the contributions of the team that performed a comprehensive
4! carbon fiber risk analysis. Overall program manager was Mr. Robert J. Huston. The tech-
nical elements of the program were directed by the following team members: Dr. Wolf
Elber (Dissemination, Redissemination and Risk Computations), Dr. Vernon L. Bell (Fiber
j Source), Mr. Richard A. Pride (Demonstration Testing), Mr. Arthur L. Newcomb (Electronic
SInstrumentation and Testing), Mr. Israel Taback (Vulnerability), and Mr. Ansel J.
Butterfield (Industrial Surveys).
15-14
Medium
(30 to 1000 Volts) Sparking or shorts Some sustained arcs
Possible fiber burn Fibur bur,.s
Transients Transients
Blown fuses Blown fuses
Stressed components Stressed components
Low damage potential Damage usually
repairable
High
(> 1000 Volts) Sparks, no sustained arcs Sustained arcs
Low voltage corona Corona
Transients Flashover
Interruptions May be severe damage
9 Dishwashers a Percolators
* Clothes washer & Food mixers
* Clothes dryer * Can openers
A * Vacuum cleaners * Portable heaters
Irons
With Failures
None
iu
15-IS
With No Failures
9 Portable heater
With Failures
o Military equipment (70) * Connector blocks
(High modulus fibers,
restricted lengths) * Quick disconnects
* Computer e Relays
a Flight director
TEST IV.-SINGLE CARBON FIBERS RELEASED FROM DUGWAY OUTDOOR FIRE TESTS
ISS-1 D-2
D-3
31.8
52.0
34.9
2.1 x 108
1.1 x 108
2.9 x 10 8
4.4
5.2
3.3
4.4
4.1
4.7
62
38
64
.19
.07
.18
I
'ji
S- 1822x1 832464 1
V
15-16
Wire, cable 1
Electrical equip. 6
Automotive fab/assy 4
I
"2!
'41
~i I, "
I +-i4
]'A
.
15-17
DC10 EXECUTIVE
OFFICE
RUDDER
(61Ibs) OFTHEPRES
IDENT
I) OFFICE
OFSCIENCE
AND
L- 0 POLICY (OSTP)
TECHNOLOGY
VEIRTI1CAL
FIN (1039Ibs)
GMETSUPPORT
AILEOS419Is HORIZONTAL
TAIL1185
IbSI
EILEVAORS
11096
bs)
EPAO
VERTICAL
FIN 110681851E
Fiber Study.
_____0
3.0
T
4 OF
~KILOGRAMS1.
SPORTING
GOODS
1.0 I
J Figure 5. Scenario of the carbon fiber
0.53NUTRA risk analysis.
R ~~-FCSISK
Figure 6. Elements of carbon fiber risk
analysis program.
iiI;I
-n
QI
Figure 3. Boron fibers released from Figure 7. Scenario for crash of civilian
aircraft crash. airliner.
15-18
SINGLE
FIBERS GUSA
SIZE: 3TO8prm DIA.. O.1TO 15mm LONG DISTPRIBUTIO
FALLRATE: 2cm/sec
DISPERSION RANGE: 0 TO>lOOkm
INVERSION LAYER
CLUSTERS OR LINT IZFULL REFLECTION
HUNDREDS OF FIBERS
FALLRATE: 10 - 20 cmlsec SOURCE<PARTIAL
DISPERSION RANGE: OTO 10km .
GROUND
STRIPS
SINGLE LAMINAE: 0.15mm THICK. VARYING
LENGTHSANDWIDTHS
FALLRATES: ! TO5m/sec
DISPERSIONRANGE:OTO 2km Figure 11. Gaussian distribution of pol-
lutant in drifting cloud, with
inversion layer and ground :e-
IMPACT FRAGMENTS[
MULTIPLELAMINATEPIECES
OCCURS ONLYIN IMMEDIATE VICINITY
Liflections.
-,---
OFCRASH/FIRE
"- WIND
DIRECTION
/ ~CLOUD)OF FIBERS
COMPOSITE BURN+
EXPLOSIVES EXPOSURE
CONTOURS
ORNIGHT
WEATHER
OVERCAST WEATHER
SUNNY
NEAR-FNIC.A IO km -A50
3 M
VELOCITY AIR FLOW
SGLOW
SIGE(FALL
.1'VELOCITY 1
" J IMPACT (DROP ORMECHANICAL) OO rdsedc CONTOUR
FOR
E= 5%OF
EMAX.
.0TWIST OR FLEX VELCIT M N--~
. -7 ,k
VBRATION10M 5km
DROP (TALL
.01 VELOCITY3.k
BURNONLY
Figure 12. Weather conditions and fiber
0 dispersion.
10 CITY
22
2. EXPOSURE OR
DOSAGE, E= CONCENTRATIONxTIME ,o1 FOOTBALL
FIELD
2 APARTMENT
:I
3. DEPOSITION,
ULJ.I: . D=
UAREA
NUMBER OF PARTICLES le 1oS
EXPOSURE.
le o0
1o0FIBER-sec/m 3 1?
9
Figure 10. Measures of carbon fiber Figure 13. Parametric plot of carbon
pollution. fiber area coverage.
15-19
FIBER 800D
FLUX,
FIBERS6000 PROBABILITY
A OF FAILURE,
-DAY63.2
2000
0 2 4 EXPOSURE
TIM, YEARS
PROBABILITY OF FAILURE 1 - e -WE/EI
E= AVERAGE
EXPOSURE TOFAILURE
Figur- 14. Extent of carbon fiber re- EXPOSURE PERCENT
FAILURES
suspension with time. "o600 1.0%
"TAO 9.5%
"r 63.2%
IT 99.9%
DiSrENSING ASPIRATOR
SENSORS
FIBER
10.0
~FIBER SKEIN-
A 8.0
)
AVERAGE
FIBER 6.0 CHOPPER
LENGTH. ASSEMBLY-
Tmm 4.0
2.0 _____0 .
I I ISTARTIC
0 1 2 3 4TARGI
YEARS TARGET
SENSOR MONITOR
RACK
Figure 15. Change in fiber lengths with RACK
ELECTRONICS &TEST
ENVIRONMENT RACK
resuspension.
I' Figure 18. NASA carbon fiber exposure
chamber.
19501
- 1960 EQUIPMENT *
HIth :MPEDANCE
HIGH PO,"R DISSIPATION * *
FILTER PERCENT HIGH MOOUL,'S FIBERS 0 TET
TRANSMISSION EATEST FIBE
FACTOR I mm INTERVAL LET
LENGTH, LI ITM
1.0 40 r-1 LIRE-RELEASED
FIBER mm 0
0.1
WIRESCREEN U MEAN=2mm 3
EQUIPMENT,
FAN-COOLED
"'' " " "
FILTERED
;] WIRE-SCRLkr4 STRIPS
OPEN TERMINAL
00 20 MEAN= 0.95mm FORM LOWER
BOUNDARY
SO00 0
0.01
FURNACE
FILTERo
01 4
--
io o6 FIBER-sec
o
O.01 5 LENGTH,
FIBER mm1 5 0 5LENGTH,
10 15 T (AVERAGE
EXPOSURE TOFAILURE).
F 3
FIBER mm
(
2 15-20 -- - - - .Ii
12
F STEREO
POWER
AMPLIFIER
(FANCOOLED
NOFILTER,
1.0 -I G NFIES
F= 1- e - ED
AE E
r I I I I
l oo 10 10,
1
lO8 0 2 4
FIBER-sec/m3
6 8X1O-
3
EXPOSURE,'.
AVERAGE FIBER-sec/m Figure 23. Probability of failure of
Figure 20. Effect of fiber length on stereo amplifiers to fire-
Fequipment vulnerabilityh released fibers in comparison
to virgin fibers.
10~ DE114
STEREO
AMPLIFIER
FIBER W/m
w 11\
RESISIANCE 14
ohms/cm 10 Figure 24. Balloon-supported "1jacob's lad-
T3 der" fire plume sampling net.
TV
COLOR HMS
t ::;A.TC TRA.NSPONER
-P 0FREQUENCY
O - 6.7PER10MILLIONOPERATIONS
103 I [0 - ABOuI"3.8PERYEAR
10 100
- FIBER-secI
0 140 SUBSTANTIAL
DAMAGE
AVERAGE
EXPOSURE, E. 3 (3mm FIBERS) FRACTION
OF I---
AEffEc OfR EICIDENTSS, TOTAL
DESTRUCTION
mM 1% 3D0-
S
Figure 21. Effect of fiber resistance on
F e equipment vulnerability.
10o
LANDINGCRUISE
STAKEOFF STATIC
TAXI
LLJ-__
15-21
a. 0 0
HH
4J
00
z 0
t33
f >
If~ 0
ui 4
~4 H
j C
$4
o p0
ti t
16-1
George S. Springer
Department of Mechanical Engineering
The University of Michigan
Ann Arbor, Michigan 48109
SUMMARY
NOMENCLATURE
:1ni Sm2
number of impacts per unit area
incubation period, number/m
2
during the
as
ae
0I
ao
stress of amplitude Ga, N
equivalent dynamic stress,
endurance limit, N m-2
ultimate tensile strength,
m
N m-2
N m-2
A nf number of impacts per unit area prior
the final erosion region, number/m2 to
7
u
average stress on the surface of the
n: number of impacts per site, dimensionless coating, N m-2
t na number of impacts pex site corresponding impedance ratio, dimensionless
to characteristic life, dimensionless
ni number of impacts per site during the in- Subscripts
cubation period, dimensionless c coating
, N fatigue life in cycles, dimensionless f fiber; or beginning of final erosion
p probability of failure, dimensionless region
P impact pressure, N m- i end of incubation period
r radial coordinate in plane
2
of surface, m m matrix
S strength parameter, N m L liquid
2
Sc strength parameter for the coating, N m a solid
Sec effective strength parameter for the sc coat-substrate interface
coating, N m- 2 Lc liquid-coat interface
t exposure time to rain, s w water
Sti incubation time, s
INTRODUCTION
1nWen liquid droplets impinge repeatedly upon a solid surface tOe stresses created by the droplets
may damage the material significantly. Fiber reinforced organic matrix composites are particularly suscep-
(t tible to such "rain erosion" damage. Therefore, in order to utilize the full potential of composite
I materials, the response of slich materials to rain erosion must be known.
16-2
The desired information on erosion damage can be obtained either through testing or through analy-
tical models. Indeed, considerable data have been generated in the past )n the erosion of composite
materials. These data provide information on the behavior of selected materials under selected conditiona
but do not describe material behavior beyond the range of the tests in which they were obtained. Alone,
the test results neither predict the behavior of materials under untested conditions, nor indicate the be-
havior of untested materials. These shortcomings of selected testing can be overcome by the use of analy-
tical models. In this paper a model ia described which can be used to estimate erosion damage of both
coated and uncoated fiber reinforced composites.
THE PROBLEM
The following problem is considered. A body, made of a composite material, moves through a rain
(Fig. 1). The impinging droplets damage the material causing pits, cracks, and weight loss. We are con-
cerned with the damage as manifested by the weight loss. Under actual conditions the weight loss generall5
varies with time as shown in Fig. 2. During the incubation period the weight loss is negligible. ror some
time after the incubation period the weight loss is nearly constant. Past this region the relationship be-
tween the weight loss and the exposure time becomes complex. Here attention is focused only on the incuba-
tion period and on the constant weight loss rate past the incubation period.
To facilitate the development of the model, the weight loss is replaced with the mass loss per unit
area m, and the time with the number of impacts per unit are n. The data are then approximated by two
straight lines (Fig. 3). Accordingly, the mass loss is
m- 0 n < ni ()
The material loss m produced by a certain number of impacts n can be calculated once the incubation
period ni and the subsequent mass loss rate o are known. Therefore, the problem is to determine ni and a.
The rain is considered to have uniform spatial distribution and to be composed of spherical droplets.
The drop diameter d, impact velocity V and the impact angle e for each droplet are taken to be the same.
The parameters n and t are then related by the expression
n
21V
6 V cosO It
td
3
(3)
The composite material is assumed to contaitt uniformly distributed, noncontiguous fibers which are
parallel to the surface. In addition, it is assumed that a) the composite is macroscopicall homogeneous,
b) locally both the matrix and the fiber are honogeneous and isotropic, and c) there is a perfect bond be-
tween the matrix and fibers and between the coating and Lhe sub..trare. Wave reflections are neglected in
uncoated composites and in the substrate of Loated composites. However, wave reflections within the coating
are taken into account.
p fvf + PmvM C 2
v is the volume fraction and E22 is the equivalent Young modulus in the direction normal to the fibers.
The subscripts m and f denote the matrix and the fiber.
A On the basis of the foregoing simplifications a model is described which provides the incubation
period and the subsequent mass loss rate. Details of this model were presented in references t and 2 and
will not be repeated here. In the following, only the method of approach is emphasized and Che results are
summarized enabling the interpretation and assessment of the results.
{i INCUBATION PERIOD
The model to he formulated is based on the idea that fatigue plays an important role in the erosion
process. Each droplet which impinges on the surface creates a stress at point B (Fig. 4). The effects of
theseof stresses
ber are takenThetonumber
stress cycles. be cumulative.
of stress The material at point B fails after it underwent a certain num-
cycles after which failure occurs is calculated by Miner's
rale (3)
+ ""+ fN al16
f22af
N1 + N
SN 2 q
!q. , I
lU-3
where fl, f2 ... represent the number of cycles the specimen is subjected to specified overstress levels
1i, 02..., and NJ, N2 '.. represent the life (in cycles) at these overstress levels, a, is a constant. A
droplet impinging on the surface at a distance r from point B creates a stress a at the point B
F(l - 2v )
o~r,0) -(7)
2
2ir
where F is a point force and v is the Poisson ratio (Ref. 1). The stress at point B fluctuates. This
time varying stress is generally replaced by an equivalent dynamic stress (4)
(o/2)((O)
a (a/2(GU)
- (8)
e au - a/2
The number of cycles for which the material at point B is subjected to a given stress between ae
and ae + dOe is equal to the number of impacts on a surface element of area rdrd
In order to perform the integration, the fatigue life N must be known as a function of che stress.
Fcr N the following approximation is introduced (Fig. 5):
(ub b2 b2
N ue) b =log0( /2 ) b-log (a /0l) (12),(12)
'e log2(au /aI lg10 (aum/aIM
For fiber reinforced composites the ultimate tensile strength ou and the endurance limit a, are calculated
by the expressions
0 oa E (13)
u um Em
E
oI- ana
i Fl~ (14)
where
40 (b -1) 1 2 2v 2v 2 + 1 + 2v1 2 1(7
Em 8 22 (11 22
m12 - 1111 12.
The parameter S represents the "strength" of the material. Thus, the result shows that the number of im-
pacts needed to initiate damage is proportional to the strength to pressure ratio. In order to extend the
range of applicability of Eq. (16) we write arbitrarily 82
n: - a1( (18)
SIN
The
(5). constants a, and
By adopting the a2same
wereconstants
determined
for bycomposite
matching materials,
Eq. (18) to
we data generated for homogeneous materials
obtain
* - (
ni . 7:x10 (19)
4 For coated materials stress waves travel inside the coating as illustrated in Fig. 6. These stress
waves were included in the model by assuming that the waves are one-dimensional and propogate normal to the
surface, and that there Is a perfect bond between the coating and the substrate. It was also assumed that
failure occurs first in the coating. The calculations including wave reflections are lengthy atAd are docu-
mented elsewhere (1,2,6). Therefore, only the result is quoted here, which is
and is the average stress on the surface of the coating. Sc is the "strength" of the coating material
as given by Eq. (17). The parameters k and 'sc as well aspF are specified in Table 1.
In Eq. (21) 2kIsc! represents the reduction in "strength" of the coating due to stress oscillations.
Note that if the coating is very thin, the strength Sc may be reduced to the point where the incubatica time
is actually shorter with the coating than without it.
The erosion rate past the incubation period is also modeled by drawing upon fatigue concepts. The
probability that a specimen will fail in fatigue between a minimum life ki and any arbitrary longer life .
may be estimated from the Weibull distribufton
p-l exp 1 (22)
where Pa is the characteristic life corresponding to a 63.2 percent failure point and 8 is a constant. For
erosion, the life may be replaced by the number of impacts n* and the amount of material lost by p. By
introducing the dimensionless parameter
m* . Psd
M (23)
a4
and by relating m* and p by a power-law-type expression m* - a 3 f , we obtain
tn* -n- a 4
md a, - - (24)
p sd. n*
p is the density of the material on which the droplets impinge and a3 and a 4 are constants. Equations (2),
and (24) give n*)a4a
a = a3 71(25)
P std3/4 (n*)a n -n*
8a i
I was postulated that a isIt a constant and is independent of a* and n: (Fig. 3). This condition is met ifi
- a (26)
!~ nsd'/4
'. we obtain
ct* a (27)
a* = a7 *)a6
i
Syhe constants a6 and n7 were evaluated by comparing Eq. (27) to data obtained with homogenous materials (5).
SByadopting the same coistants for composite materials, a* becomes
. ( n* ) 0 7
RESULTS
Equation (28) appli.es both the coated and uncoated composites. The expressions for calculating the
incubation period and the rate of mass loss are summarized in Reference I. The results are compared to data
in Figs. 7 and 8. As can be seen, there is excellent agreement between the measured and calculated n and
a* values over a wde range of conditions. This agreement lends support to the validity of the model and
suggests that the model may be used to analyze and pred~ct Lroslon damage and to guide in material develop-
ment. The model also provides a means for comparing the response of different materials to erosion as illus-
trated by the corparisons presented below.
forthIn Figs. 9-11 the incubation period and the rate of mass loss are presented for three metals and
for three fiber reinforced composites. The incubation periods of graphite-epoxy and glass-polyester compo-
sites are several orders of magnitude lower than for metals. Only the incubation period of boron-epoxy com-
posites
ei comparable to that of one of the metals (aluminum). However, even boron-epoxy has an incubation
Speriod which is alout four orders of magnitude lowec than steel. The mass loss rates of composttas are alau
lower than those of metals by several orders of magnitude.
(Figs.The impact velocity affects significantly both the Incubation period and the rate of mass loss
(Figs. 9, 11). At impact velocities (Vcos 0) below 200 m/s the incubation period for graphite-epoxy compo-
sites is about an hour. The incubation period decreases rapidly with impact velocity. At 400 m/s the incu-
bation period of graphite-epoxy is only about one minute. Consequently, the material should be protected,
particularly at high impact velocities. One convenient method of protection is by the application of a
coating. The appropriate coating material and coating thickness, as well as the protection provided by the
"coating, may be calculated by the expressions given in Table 2. The results of sample calculations are
16-5
shown in Fig. 12. The result3 in this figure illustrate that the coating must be selected carefully. Some
coatingb do not provde protection and may even reduce the incubation time. For example, the many stress
oscillations in the teflon coating on graphite-epoxy weaken the coating to such an extent that the incuba-
tion period for the coating (teflon) is actually shorter than it is for the substrate (graphite-epoxy).
CONCLUDING REMARKS
The erosion model presented here -as developed on the basis of physica concepts and experimental
observations. Although there are a number of approximatior- ind simplifications included in this model,
the results agree well with existing data. Nevertheless, the limitation introduced by the approximations
should always be borne in mind when applying the results to rew situations.
REFERENCES
1. Springer, G.S., Erosion by. Ljguid Impact, New York: J. Wiley and Sons, 1976.
2. Springer, G.S. and Yang, C.I., "A model for the rain erosion of fiber reinforced composites,"
AIAA Journal, 13, 1975, 877-883.
3. Miner, M.A., "Cumulative damage in fatigue," Journal of Applie Mechanic., Trans. ASME, 12,
1945, A159-A164.
4. Juvinall, R.C., Stress, Strain and Strength, New York: McGraw-Hill Book Company, 1967.
5. Springer, G.S. and Baxi, C.B., "A model for rail: erosion of homogeneous materials," in Erosion,
Wear and Interfaces with Corrouion, ASTM STP 567, Society for Testing and Materials, 1974,
106-122.
6. Springer, C.S., Yang, C.I.. and Larsen, P.S., "Rain erosion of coated materials," Journal of
Composite Materials, 8, 1974, 229-250.
7. Lapp, R.R., Stutzman, R.H., and Wahl, N.E., "Summary report on the rain erosion of aircraft
materials," 1956, WADC Technical Report 53-18.,, Part 3.
8. Lapp, R.R., Thorpe, D.H., Stutzman, R.A., and Wahl, N.E., "The study of erosion of aircraft
materials at high speeds in rain," 1958. WADC Technical Report 53-185, Part 4.
9. Schmitt, G.F., Jr., and Krabill, A.H., "Velocity erosion relationships of materials in rain at
supersonic speeds," AFML-TR-70-44, 197d.
10. Schmitt, G.F., Jr., "Rait erosion behavior of graphite and boron-fiber reinforced epoxy compo-
site materials, AFML-TR-70-316, 1971.
11. Schmitt. G.F., Jr., "Material parameters that govern the rain erosion behavior of polymeric
coatings and composties at subsonic velocities," AFML-TR-71-197, 1971.
12. Lapp, R.R., Stutzman, R.H., and Wahl, N.E., "A study of the raiui ruols uof piasti.cs and
metals," 1955, WADC Technical Report 53-185, Part II.
v v
MOM Ma
P e
E Incubation 1SteadyWe~gh lFino! Eosion Region
Fig. 1 Droplet Impingements on un. and LossRole
RFepad
coated fiber-rat inforced composites Wdt I I
i I I
II
i ~U Ii
000
0
Fig.NUMBE
Fig. 3
3UNITIMPCT
~Representation
fi_ _ _ _
AREA,; moe
ofthmoe
EXPOSURETIME, I
Imping Dropie /
Poll t Force ^ A:-"0 0 0" /
F. -.d P 0
40 0 0 * 0
0/ d*'/ A 7/ y 0,0**
0d 0 iomn
/IU o S
SUBSTRATE04t /Cc
/ o hCCC0 t holtc
S(b
2
or IN/CC tI 11C/CC
0'
CY
cr' Od'o0
'LOI -- W-A7.'SUSRAE
II
Io I
~10
| .. .. I, , , ,.. n , L ,I
.~~. zo~
-
AP
L C0 Lg
s
Fi .o i
k Fig. 7 Comparison between calculated and measured
2 incubation period and rate of mass loss
f, for uncoated composites. Data are from
:, References (7) - (11)
sit L'-
--
[..i
__
~ .-- ~ .-- .--. ---
___to
- P-
ItI
16-7
104 o
o0
lo0
00 07 ...
'C
c
Id
102
10 1016
j 3 lass/Polyester
U) GE
B/-
on 105
Stee
w
jI~ 1.0
I(5-
z
0-6
Bro n/Epoxy
~ 010 '~rophite/Epoxy
ZGloss/Polyester
1012
to
10p
zL 101
01
10
z
10) - ., 1-.. -
200 400 600 800 1000
IMPACT VELOCITY, VcosG(m/s)
Fig. i1 Rate of mass loss of different uncoated
materials calculated by the model. Rain
intensity 25 mm/hr, drop diameter 2 mm.
Io
106
ALUMINUM COATING
w
i
A A~ TEFLON COATING
100o0
j
ki, 16-9
TALE 1 Parameters used in the calculation of the strength of coated composites. The units to be used
for each parameter are given in the "Nomenclature."
4
Lc ZL - Zc Dimensionless
A
the coat-liquid 00 r
interface scL sc 1+ ty
*1
.i "
41i
R44I
RECORDER'S REPORT
Session IV - Physical Hazards
by
Renato BARBONI
Scuola d'ingegneria Aerospaziale
dell'Universita di Roma
Istituto di Teenologia Aerospaziale
Via Eudossiana, 16
00184 Roma - Italy
The aim of Session IV was to examine the physical hazards related to the use of composite
materials. Three fundamental aspects were taken into account:
1. Effect of lightning on aircraft and the protection requirements.
2. The risks of accidental release of carbon fibers from burning of crashed
commercial airlines.
3. The material damage due to repeated impingements of rain drops.
As far as structural damage due to lightning is concerned, the results of two experi-
mental investigations which pointed out the thermal and the high current density effects
; were presented.
reaction Both papers
of composites agree with
to lightning the basic
strikes point of
in a manner view that,
different frominconventional
spite of the
materials, the resultant damage is by no means alarming. Nevertheless, I would note
some interesting considerations that appear from the papers presented by Rouchon and
Schneider:
i) the importance of honeycomb core material for sandwich structures
ii) the sensitivity of some types of joints to lightning
The third paper, by Bell, was concerned with experimental and analytical results on the
potential damage to electrical and electronic equipment and the risk to the community
due to accidental release of carbon fibers. The graphite fiber risk analysis scenario
assumed carbon fibers to originate principally from fires from czashed commercial
aircraft with carbon composite parts. The unusual situation for experimental research
was the absence of actual experience because fortunately, up to now, crashes of com-
mercial aircraft with any carbon composites in their structure have not occurred. Of
course, simulated experiments have been carried out and interesting results on the
} burning, dissemination and redissemination of graphite fiber obtained. The results
indicate that:
1.
1 about the vulnerability or susceptibility of electrical and electron
equipment,many of them were invulnerable.
2. no risk to the community is to be expected from the use of carbon fibers
in civil aviation.
3. the chances of national losses reaching significant levels are extremely small.
4. furthermore, the potential shock hazard from carbon fibers is insignificant.
In conclusion, no problems are posed to the community by an extensive utilization of
carbon fiber and programs to develop alternate materials specifically to overcome that
perceivc.d hazard are not necessary.
The response of composite materials to liquid droplets erosion is examined in the last
paper of Session IV. Springer proposes an analytical model based on the idea that
fatigue plays the dominant role in the erosion process. The incubation period and the
Ssubsequent rate of mass loss have been chosen as the fundamental patemeters in order to
evaluate the real life of the material. Theoretical considerhtions lead to very simple
formulae to predict the number of impacts needed to initiate damage, and the erosion
rate. Comparisons between theoretical results and experimental data excerpted from
~ other authors seem to certify the validity of the model. Unfortunately the application
of the model to graphite-epoxy composites gives incubation periods of several orders of
magnitude lower than for metals. This means, for example, that the real life of CFRP
R4-2
wiinutes.
utilized in a civil aeroplane exposed to rain is only of the oxder of several
dangerous
The conclusion from the papers presented in Session IV is that the most
phenomenon, rain.
physical hazard is for CFRP and is di'e to the most natural physical
I ".
A .
17-1
SUMMARY
1. INTRilETIOij
During its service life the helicopter experiences a severe high-cycle fatigue environment, the total
number of cycles being, typically, of the order of 109. This condition is particularly demanding for the
rotor blades in which stress levels are necessarily high due to thu strict constraints which exist on
geometry, aeroelasticity and weight. To design a blade which will endure this fatigue environment is
an objective which, with metals, is extremely difficult to achieve.
To overcome this problem helicopter rotor blades are now being produced from fibre-reinforced compo-
sites. At Westland Helicopters glass fibre in epoxy resin f.slargely employed with selective nse of
carbon fibre for additional stiffening where necessary. These materials have the high fatigue strength-
to-density ratios which when properly used in rotor blade design allow components with unlimited fatigu?
life to be produced.
The use of composite materials also permits the advanced blade shapes now being demawded by the
rotor system designer to be produced accurately and economically, and hence large improvements in rotor
performance can be achieved. Additionally, the use of caposite materials allcws blades to be designed
which exhibit a high level of damage tolerance. Blades manufactured with glass-reinforced epoxy spars
show extremely low rates of damage propagation which contributes to improv-1 safety. One of the most
t :exciting potential advantages of composite rotor blades lies in the ability to manipulate, to a high
degree, the blades' elastic properties and hence to tailor blade designs for optimum dynamic and aero-
elastic performance and minimum vibration levels.
This paper describes recent work performed by Westland Helicopters on composite tail and main rotor
blades. The paper concentrates on the performance of blades for the Westland Sea King and WG.30 helicopters
. under the fatigue environment in both rig sad flight tests.
2. BLADE CONSTRUCTION
Blade construction is, typically, of a hollow 'D'.spar composed largely of unidirectional glass-
reinforced plastic with internal and external woven wraps. The trailing-edge skins are of spanwise uni-
directional and angle plies. In tail rotor blades built to date these trailing-edge skins have been
* "manufactured in carbon fibre/epoxy and supported by a rigid polymethacrylimide foam filling. In main
blades, however, the trailing-edge sxins may be either carbon or glass fibre/epoxy (or a hybrid of the
I' two) on a 'Nomex' honeycomb core. A bolted joint is used at the root-end in all cases with local rein-
forcement of the spar. There is provision for the fitting of a heater mat on main blades if this is
recuired for de-icing purposes. A titanium erosion shield provides the necessary leading-edge protection,
and tip weights and a tip cap are installed at the outer end of the blades.
The construction of a tail rotor blade is detailed in Fig.1. This blade is designed for the Sea King
and is shown in its ready-for-flight state in Fig.2. A cross-sectional view of a Sea King main rotor
blade is shown in Fig.3.
17-2
3. TESTING PHILOSOPHY
The approaca adopted for the clearance of composite rotor blades for flight incorporates two series
of tests. The first series is performed upon what is known as 'Structural Elements'. These tests provide
S-N curve shapes and statistical data on the scattnr characteristics of the structure/material in various
critical areas of the blade. It is therefore important that structural elements are either cut directly
from actual blades or manufactured by the means, and from the materials, of the blade itself. It is from
this series of tests that the materials scatter factors employed in the design are validated, and there-
Mean strength values, however, are not sought from these tests. These are derived instead from the
second series of tests which are performed on a liaited number of full-size blades. The root-end and out-
board aerofoil sections of the blade are tested by separate specimens. The fatigue testing involves the
application of programmed series of loading blocks designed to correspond closely to actual flight condi-
tions.
t
In this way the costly and lengthy process of full-size blade testing is minimised. The two series
of tests examine both the static and fatigue performance of the specimens. Full-size specimens are also
used to assess the damage propagation characteristics and to investigate the effects of extreme conditions
of temperature and humidity (both natural and artificial).
The positions of these areas are shown in Fig.4. In most cases so- development work has proved
*1 necessary on the design of specimens to ensure that a representative failure takes place in the test
section. The lay-up of the Sea King tail rotor blade structural elements is shown in Table 1. A tensile
& . loading was used for these tests. In the fatigue case a 1 .1P !P form was adopted. The failure modes of
the different structural elements for this blade are described below:
- '4i)
I ~ Spar
The failure modes obtained for three of the spar specimens are shown in Fig.5. This type of
failure is similar to that obtained in the static case with cracking occurring along and across
the 450 cross-plied carbon and woven glass layers, together with pull-out of the unidirectional
glass material.
ii) Trailing-Edge Skins
) This specimen consisted largely of epoxy/carbon fibre. Specimen development proved more
4difficult
ki
in this case and a more violent tensile Jailure .tode occurred, causing severe damage
to the element.
iii) Root-Ehd
In the static tests failure occurred in a lug tension moae with some local pull-out of
unidirectional glass. In the fatigue case the failure modes differed from those seen under the
static tests in that much less fibre pull-out occurred. Four of these fatigue specimens can be
seen in Fig.6. Although a good deal of damage to the flat surfaces around the holes can be
"1 seen, there is no evidence of failure origins between the bush and the bore of the holes due to
fretting effects.
17-3
iv) Transition
These elements gave both static and fatigue failure modes wldch proved very similar to those
of the spar elements.
The results from the above structural element tests performed for the Sea King tail rotor blade are
shown in Figs. 7-10. A summary of the values calculated for material scatter factors is given in Table 2.
These factors, which are for use when two full-size specimens are tested, are encouragingly low and com-
pare favourably with those assumed in the original design (also given in Table 2). If more than two full-
size specimens are tested these will, of course, reduce.
The variable-amplitude structural element fatigue tests for this blade are in progress at present.
(The results from these tests were not required for the initial 50-hour demonstrator clearance).
Also in progress at prebont is the structural element testing for the Sea King main rotor blade
Demonstrator. At the timLe of writing, inforMation is complete for the trailing-edge skin elements only.
Again encouraging results have been obtained. A static coefficient of vwriaticn for the population
(7 specimens) of 0.046 gives a two-specimen material test factor of 1.07. In constant amplitude fatigue
a two-specimen material scatter factor of 1.37 has been demonstrated.
Fig.11 shows one of the rigs used for the Sea King tail rotor blade tests. The static tests
relate to two load cases. One case represents the peak flapping and lag bending moments and torsional
loads occurring in the worst flight condition together with centrifugal loads appropriate to normal rotor
speed. The other case represents the effect of rotor overspeed and consists of a peak centrifugal force
together with nominal values of flap, lag and torsion.
The fatigue tests are carried out under a programme of vibratory flap, lag and torsional loads
together with steady C.F. loads in a series of ascending - descending blocks. Between each block is
interspersed a programme of once per flight applications and relaxations of centrifugal loads together
with a flap and lag moment and torsional load.
For the Sea King tail rotor blade the two static specimens for the outboard case achievwd 105% and
110% of design ultimate load, respectively. This load includes a 1.5 aircraft ultimate factor and an
A assumed 1.5 materials scatter factor. Fig.12 shows that the failure mode is one of wrinkling of the
trailing edge skin. This is caused by the lag bending component generated due to the fact that the
"centrifugal load path is offset from the blade neutral axis at the outer sections of the blade. In the
case of the root-end static specimens, failure occurred at 100% and 106% respectively, of the design
ultimate load.
Fatigue testing of an early standard of Sea King tail rotor blade root-end specimens indicated a
i weakness in the blade spar near to the erosion shield termination which limited the life of this standard
of blade to about 87 hours. Design and manufacturing changes were made at the root-end and a new standard
of blade was produced. After the fatigue testing at fully-factored, programmed loads of two root-end speci-
mens to this standard no failure occurred and hence unlimited fatigue life has been shown.
A photograph of one of these specimens after it had finished its complete, fully factored, fatigue
testing programme is shown in Fig.13. A crack of about 20 mm in length can be seen in the adhesive near
the erosion shield but this crack does not extend to the load-carrying structure of the blade. At the
root end, de-bonding of the bushes had led to slight fretting, and some scorching of the surface in con-
tact with the tdummy' blade cuff had occurred. Additional fatigue loading of a similar blade specimen
(which had already sustained the complete fully-factored fatigue testing programme) at first 120% and then
140% of the fully-factored loads, eventually lead to a wear/bearing failure in the composite material at
the bolt holes.
"TheFatigue testing of the outboard specimens completed 107 cycles of fully-factored, programmed loading.
The titanium erosion shield failed during this testing but the blade itself has remained unfailed and has
demons~rated an unlimited fatigue life.
17.4
Damage propagation testing has been performed on (known) sub-standard blades with initi-L! cr.'ks
of approximately 10 mm in the spar at the root-end high bending load position. After being siLjected to
a full programme of flight loads no damage propagation or loss of stiffness could be detec ad.
Work on the fatigue testing of full-scale composite main rotor blade specimens is under way at
present and results to date have shown that the excellent fatigue and damage propagation properties
obtained writh tail rotor blades can also be achieved with main rotor blades.
In the forrer catagary the approach is to subject a total of 20 complete blades under load to
severe natural environments as follows;
After exposure of up to 5 years these blades will be returned to W.H.L. and converted into equal
numbers of root-end and outboard test specimens. The blade specimens will then be tested under fatigue
and static loads in exactly the same manner as described in the previous section. Stiffness and strength
results will then be compared with those obtained from unexposed blades.
Artifici&I ageing work has been in progress for some time on coupon specimens. The objective of
this activity is to investigate the temperature -humidity-thickness-time relationships and define the
conditions under which full-scale artificial ageing tests should be conducted. Twelve complete composite
tail rotor blades are allocated for this full-scale work. After exposure, 6 root-end and 6 outboard test
specimens will be made from these blades. Static and fatigue testlug will then be performed for direct
comparison which tests performed on naturally aged blades. The development of a reliable accelerated
ageing technique is clearly essential for the economic and rapid appraisal of environmental effects on
composite rotor blades.
coverage of the flight envelope has been achieved with blade loads proving to be close to those calculated.
Blade lives derived from rig tests and flight measurements are again virtually unlimited.
6. CONCLUSIONS
4 The rig and flight test programmes performed on composite rotor blades have shown that blades of high
structural integrity can be designed and manufactured in composite materials. It has been demonstrated
that unlimited fatigue lives can ,be achieved and that advanced aerofoils can be produced accurately and
economically thus making possible significant performance improvements.
!.I,
.The
, structural element teits have shown encouragingly low material scatter in both static and fatigue
cases. It seems that static factors of about 1.2 and fatigue factors of about 1.5 could eventually be used
to achieve a level of safety comparable with that of metal blades. Indeed, these factors are themselves
I fY-" -
17-5
Damage propagation testing performed on composite blades to date has shown that, under flight load
levels, propagation does not occur from significant defects (about 10 rm) in the G.R.P. blade spar. This
fact can contribute significantly to improving helicopter safety through the use of composite rotor
blades.
Work at Westland Helicopters is now concentrating on the gathering of further test data for the
composite tail rotor blades on WO.30 and composite main rotor blades on Sea King and WG.31. This
involves, in particular, further rig testing of root-end and outboard specimens and the collation lf the
information being generated from blades subjected to extremes of temperature and humidity.
ACKNOWLEDGemENtS
Much of the work reported in this paper was funded by the Ministry of Defence under the Helicopter
Technology Demonstrator Programme.
Ii
iil
17-6
Table 1 Lay-Up Summary for Sea King Tail Rotor llade Structural Elements
CGBA-GEIGY Fibredux
EROSIOIJSTRIP
TIP CAP ANDEROSIONSTRIP EARTIIEU. TO AIRCRAFT
L- - --SCO~
- -
o
~ aFig.2 Completed Sea K~ing Composite Tail Rotor Blade
.Am.
17-8
SEA KING
-I
' ! /
ROOT END ATTACHMENT TRANSITION AEROFOIL SPAR AEROFOIL TRAILING
iI
SEA KING
1..
-~J
~alaar
4- 7M- 7,7 .7 -7 7
LNi
50. mIm
4i. alRtrBaeSa osatAmltd aiu tutrlEeet
r~ 7
RAA
i~NO
17-11
-z &
Ii
IkID 0U
w
<(00
us__ _
U) cc
-
00
ClC)
0.
N %D
I ~E-
4o
OD C-
.2 j -
< I W< LU
--- 4.--
~~- -L -- LL._C~D
U))c- o
H.0i
-
c a
V) I
17.12
43
00
0 44c
th
*SC - n
IZU~
(r <
'ID
F~D~oco
& ~W~fl(0
z 02
z0 uZ
j M -ILt
V)b LII
0
m
*0 t
>. 0
77A
17.13
-v ~~iU
a z
0-
P.
S2
cr U)
000
_ _ _ -4)
U) I F1II
w0
C,,
r= LL
< <Ly 'Cn
E~
E 0
W a-
03 m
w 0
ZZ C
UJ 'o to
z0 Z
U)ca 2f o
z h
0..
0u w
U) (AN
zz
AZ
cc
___________________
17-14
44
0-0
01-W0
*U 0 ( 4,
0 40
4)
4)
60
C!
wCc,
c 0
LL W-
;Woz
Z
5 3
(p I~z
cc
0 w
or,.
~ Icc
17-15
AM
! I ,A
JFig.12 Typical Sea King Tail Rotor Blade Outboard Failure Mode (Static)
17-16
-~- -- ~
w~
"4 - z-w V
C1 A,,
IFig.13 Tail Rotor Blade Root-Ftid Test Specimen After Completion of Fatigue loading
K N
WIr
-----------
-Mv
Fig.lh Westlend Sea King Composite Tail Rotor Blade Flight Installation
17-'77
Ii
S230
I ,i
I *
- I
18-1
by
K. Brunsch
Messerschmitt-B81kow- Blohm GmbH
Postfach 801140
8000 MUnchen 80, Germany
SUMMARY
By the end of 1979 more than 400 light helicopters BO-105 had been in service both
for civil and military operations accumulating approximately 500 003 hours of flight.
Some of the helicopters have exceeded 5000 flight hours. The BO-105 being the first pro-
duction helicopter with a hingeless type main and a teetering type tailrotor with fiber-
glass rotorblades makes the service experience with the hingeless rotor and the GRC
rotorblades unique.
The service experience accumulated with the GRC rotorblades is presented. A com-
parison of fullscale fatigue test results with new blades and blades after 4000 hours
of flight is given, also a comparison of coupon test results, coupons cut out of blades
before and after some thousands of service hours.
A general comparison of GRC and metalrotorblades is made. Some recommendations
resulting out of the service experience conclude the presentation.
INTRODUCTION
When the BO-105 started operational service nearly ten years ago it was the first
production helicopter fitted with composite (GRC) rotorblades both for main- and tail-
rotor. This pioneering event - use of GRC for the rotorblades of a mass produced
helicopter - was result of R + D efforts on composite rotorblades starting as early
as 1958, with the first blades flown in 1962. Within that R + D work the unique combi-
nations of low stiffness and excellent fatigue strength owing to 1lassfibre composite
became evident.
ii not only, it also was pioneering in system simplyfication by the application of new
materials - drastically reducing number of parts and maintenance costs (Fig. 1).
Today it can be said, that the enormous courage of the MBB management, when
taking the risk to decide for production of the BO-105 has paid off well. More than
400 BO-105 helicopters are in service worldwide.
Most of them fly civil missions, supply to German army started in 1979. The
wide field of operation and experience covers missions like off shore, utility, rescue,
police, executive, lighthouse supply and sco, t, antitank and L:)H in military operations.
The 400 BO-105 flying in 28 countries have iccumulated approximptely 500 000 hours
of flighttime, equivalent to some 2 millions of mainrotorblade hours and 1 million
tailrotorblade hours. Blades have been in service since 9 years, the maximum hours of
flight blades just approaching the 6000 hours i-ark, which is half of TBR. Flight envi-
ronment ranges from -40 0 C to more than +45*C, from dry desert areas to the 100%
humidity, hot climate of Gulf of Mexico.
The experience with the BO-105 GRC-rotorblades is of great interest, because of
the many composite rotorblades being developped or starting production just know.
Aside of the benefits those projects might draw the excellent structural experience
should spur any projects trying to develope bearingless rotorsystems. All well known
4 helicopter companies like AS, Bell, Boeing, MBB, just to name someones make R + D
efforts in this field (Fig. 2).
DESIGN OF ROTORBLADES
1 .The design of the mainrotorblade is simple - rectangular plan form (Fig. 3)
with an inboard cut out and constant NACA 23012 mod. airfoil. With reference to the ex-
perience to be presented later only the bud joints of the erosion protection should be
kept in mind. The crossection of the Airfoil (Fig. 4) shows the structural and semi-
structural components. The skin of the blade is built up of fabric with +450 fibre
orientation to provide torsional stiffness mainly. To make the torsion box continuous
18-2
all around the airfoil there is a broad bond-line at the trailing edge. The geometry
of the transition area airfoil to tab was developped to improve stall characteristics
of the airfoil. At the leading edge the lower and the upper skin are connected via the
erosion protection strip , joined to the blade by a secondary bond. At the inboard area
with its cut out the torsion box is made of GRC only. After curing of the blade two
U-shaped shells are bonded to the shin.
The C-spar, made of unidirectional E-glass provides tensile strength and contri-
butes to chord- and flapwise stiffness. The foam core machined out of plates of modified
PVC does stabilize the skin and paiticipates in shear load transfer between upper and
lower skin. The fabric wrapped around the cg balancing mass made out of lead also im-
proves shear load transfer.
The mainrotorblade is attached to the hub by a titanium fitting. The mainly GRC-
uni-structures hoese-shoe shape completely is embedded in the fitting. To compensate for
manufacturing tolerances the groove in the fitting is slightly oversized with reference
to the toblade The gap between Oitting and blade is filled with an adhesive (Fig. 5).
Prior that root.
the fitting is thoroughly sprayed with a release to avoid any bonding
blade to fitting. The fitting is attached to the hub by the mainbolt which transfers
centrifugal load and flapwise bending and by the secondary bolt contributing to chord-
wise bending moment transfer only.
The design of the tailrotorblade is like that of the mainrotorblade rather simple.
The planform again is rectangular (Fig. 6), the airfoil is NACA 0012 mod. all along the
blade span. The blade is attached to the teetering type rotorhub by two bolts. The uni-
directional GRC does circumvent the bolts of the tailrotor attachment. The internal
structure is like that of the mainrotorblade (Fig. 7).
MANUFACTURING TECHNOLOGY
It is the authors opinion that the usefullness in service of a composite component
severely is influenced by manufacturing technology .nd -quality. Therefore the manufac-
turing technology of the blades is described briefly. Tools both for main- and tailrotor
are female moulds for lower and upper side (Fig. 8). Moulds are NC-milled out of
stretched Aluminium. Aluminium was selected for ease of machininr, good thermal conduc-
tivity and coefficient of thermal expansion being higher than that of GRC.
By the later heating up of the mould tends to improve f.bre alignment rather than
to buckle the laminate. Blade production starts with a wet ha.d-lay-up of fabric sheets.
The mass of each set of fabric sheets cut for one blade is noticed before. For an
exactly specified number of sheets always the same amount of resin is to be used. To
control that the total mass of resin to be used for skin impregnation is split into many
fractions, eg like tabled below f6r Main Rotor Blade (MRB).
Everymix of resin is reported with reference to total mass, mass of resin and
mass of hardener although an automated dosing unit is used.
To provide controlability of fibre alignment- the skins fibreorientation is
+45* to blade span- the fabric is fitted with tracer fibres.
To ease impregnation moulds are heated to 43 + 2 0 C. As soon as the skin laminate
is completed the spars are built up out of roving tipes. A special technology was
developed tc ensure very hiqh roving tape quality. Each roving tape consists of 18
rovings with 768 tex (Fig. 9). The set of bobbins (Fig. 10) is put toqether in such
a way, that the mass per unit of length of the dry roving tape is always the same.
ki
18-3
Experience is, that tne mass per unit of length nearly ever is constant for all
the material on one bobbin.
Nevertheless eac', start of the impregnation unit is preceeded by a test to define
the mass of one meter of dry roving tape. The roving tape is pulled over a drum which
dives into a resin conzalner to about 40% of its diameter (Fig. 11). The drum picks up
a resin film part of which attaches to the roving tape. Then the roving tape goes
through an oven 55 0 C in thermal temperature. Thereby wetting and netting resin to glass
is improved consideiably (Fig. 12). After that excessive resin - approximately 50% of
that picked up - are squeezed out by an adjustable fixture (Fig. 13). The roving tape
then is pulled thro'igh nozzles intended to form the final geometry of the tape only
but not to squeeze out resin. Both tapes for main- and tailrotorblade have the same
cross-sectional area but different geometry.
Impregnation is done discontinuously,tapes with slightly more than twice of the
length of the MRB are produced. Of any tape a dual knife (Fig. 14) cuts a constant
length specimen dropping on a scale without being touched.The scale indicates and re-
cords individual tape spccimen mass and accumulated tape specimen mass. With the dry
roving maus being constant always the scale gives an exact information on resin content.
Fibre content is controlled to 73,8% per weight.
The roving tapes are laminated onto the skin already in the moulds starting from the
tip of the blade to the inboard attachment area, around the fitting (MRB) or attachment
bushings Tail Rotor Blade (TRB) and back to the tip again.
Afterdeposition-of 20, 40, 60 and 76 tapes per female mould (MRB) templates are used
to control for proper spar geometry.
60 additional roving teoes are laminated to the MRB spar in the attachment area.
When the spar production is complptcd a core machined out of modified PVC foam is inser-
ted to the lower mould, also a ropc of lead for chordwise cg balancing.
The "upper" mould is lifted by crane, rutated for 1800 and carefully lowered to the
other one. The two haf moulds are pressed to one another by large bolts and curing
temperature becomes appli3d. After curing the MRB is hung up at the attachment loop and
allowed to relaxe from thermal stresses induced by curing for 6 weeks.
Production is continued with bonding of the erosion protective strip and of the
GRC shells in the attachment arcL to the alade and with mating the fitting (MRB) to
the attachment lug. The point applied then paiticularly is used for spanwise balancing.
After final static balancing the blades are ready for dynamic balancing on the whirl
tower.
Following this brief description of the BO-105 GRC-rotorblades the service ex-
perience gathered with them will be described.
Strength Properties
The application of GRC for a hingeless type rotor was quite innovative when MBB
started R + D on this system. Excessive material t-esting, especially fatigue testing
was necessary to evaluate material properties. Specimen testing was done with specimen
cut out of blades only. The tooling built up for experimental blade production allowed
for manufacturing of better composite material than the equipment Available in the la-
boratories. Some thousands of spar material Lpecimen have been tested to establish ten-
sile, flexural and shear fatigue strength data. With respect to service influence on
material strength the best information can be drawn from comparison of those datas with
th..t produced from specimen which have been cut out of blades that had been in service.
il Specimen have been cut out of blades that had accumulated up 6o some 3000 hours within
5 yea s and fatigue tested. Test result indicate a shear strength reduction of 1,8 %
at 10 cycles (Fig. 15).
F]exural fatigue strength appears to be slightly higher than that evaluated with speci-
"men cut out of new blades (Fig. 16).
As there is no physical explanation the later result is considexed to caused by
a limited number of specimen. Generally can be concluded, that there is no adverse
influence on spar material strength. Taking in account the load cycles the blade has
accumulated (Fig. 17) and adding the representative number of cycles to those applied
with specimen testing shear fatigue strength also is like that of neu blades spar
A i material.
Service influence on blade skin fatigue strength vas investigated by component
testing only. Therefore the number of test results is very small. The reason for compo-
nent testing was that all efforts to develope an usefull skin test specimen (to be cut
ot the blades) had failed due to the curvature of the blade surface and the assymetry
of the desicn of the skin.(Difference of resin content eg because of foam core adjacence
"at skin at "ne side and the metal mould at the other). Two types of fatigue tests have
been performed - flapwise bending in a resonant beam bending mode (Fig. 18) ie without
simuation of CF and a tension torsion test (Fig. 19), where torsional and CF loading
was simulated. Test results again indicate, that there is no adverse influence on skin
&V
18-4
fatigue strength.
Finally the mosf- complex components of the mainrotorblade, the attachment areas,
were fatigue tested too. Fatigue test results again demonstrated that here is no ad-
verse influence on the fatique strength of the blade material composed and produced as
described before (Fig. 20).
S On
properties handreference
the otherwith the earlyof days
skin. ininfluence
significant
there wasto a blade service environment
of BO-105 on material
the inboard area
(the so called swan neck) was coated with a transparent Pu-paint to allow for visual
detection of any deterioration that might show up. Unfortunately deteriorations became
evident very soon. The UV part of solar radiation penetrated the Pu coating and severely
damaged the epoxy resin at the interface to the Pu-laquer. With some blades excessive
peel off of the PU-coating occured (Fig. 21).
ELASTIC PROPERTIES
Aside of strength prcoerties service influence on elastic constants are may be
of even greater interest with composit materials used for rotorblades. Rotorblades are
designed extremely carefully to avoid any vibration caused by the exciting frequencies
(mainrotor, tailrotor, number of blades, tailboom natural frequency and so on).
The manifold of exciting frequencies, -especially higher modes - does incorporate the
risk that even minor changes in frequency result in unacceptable vibration of the heli-
copter caused by (partial) resonance. Determination of changes in elastic properties
has been done more or less by three ways - mainrotor blades have been flown in excess
of 6000 hours without any adverse influence on vibratory characteristics of the heli-
copter. That means that non of the "eigen" modes had changed to come close to an
exciting frequency. Second a pair of blades that had exceeded 6000 service hours success-
fully has been mated with a pair of new blades. Again no adverse influence on helicopter
vibration showed up.
Finally a lot of frequency testing has been done. In the early production phase
every MRB was tested for flap- and chordwise bending frequencies 1st and 2nd mode in a
free-free test set up (Fig. 22). Again the best information concerning service influence
on dynamic properties can be drawn from comparison of frequency test results of new
blades and of blades after service. The results presented in Fig. 22 indicate that there
is no adverse influence detectable. An exact number of new blades tested is not given
because any 100th new MRB is tested for quality control. Thereby the number of blades
tested steadily does increase. Tests in a clamped-free configuration also have been
done. Due to the BO-105 specific attachment design these test showed considerable
scatter, also major differences in test results were evaluated when changing from a
set up with horizontal chordline to a set up with vertical chordline. Therefore this
testing is considered not to provide valid information. Aside of problems with testing
there is another fact making the type of test suspicious. As described before there
is a layer of filler material in between the titanium fitting and the GRC blade. This
layer initially was made out of AW 106, a special product of Ciba.
This material performed pretty well in component testing but was not acceptable in
service, due to severe deterioration (Fig. 23) caused by hot humid environment. The
filler material was replaced by a product with trade name AV 138 M and the problem
seems to be overcome. Nevertheless minor abrasion at the outboard position of the
attdchment lug still does occur. Wi.th Centrifugal Force (CF) abrasion does not influence
natural frequencies. When fatigue testing blade root components with load well above
max. manual load considerable wear of the filler materials was produced. Without re-
pairing the filler naterial frequency testing with CF was done. Test results were the
same with new components and components with the filler material worn off the outboard
areas of the attachment lug.
I Without CF - as in testing with a complete blade - the influence is remarkable. There-
fore a repair of the filler is to be done prior to testing bringing natural frequencies
back to the band known from new blades (Fig. 24).
I i1
18-5
I Shifting the removal of the PU coating from one shop (well trained with GRC-
blades) to another one, experierced with metal blades only resulted in more than a dozen
of TRBS made scrap by blade skin abrasion.
Such thingsof course do not positively influence maintenance cost. But the increase of
maintenance cost making abraded blades scrap is the by far minor a problem compared
to that an abraded blade might be taken in service again. Especially with the tailrotor-
blades blade skin abrasion significantly could reduce safety of flight. The operators
of the B0-105 evidently do know about that. Up to now no accident/incident is known which
was caused by rotorblade failure due to GRC abrasion. On the other hana there have been
sent quite a lot of blades to MBB to clearify wether the coating had been removed accu-
rately or what to be done with the blades renew painting only, repair of GRC or scrap.
Most of the blades went back to service again.
-'I:a
18-6
stopped after simulation of 5 MRB lifes equivalent to 6 x 106 cycles. Towards the end
of testing the delaminated area had approximately trippled compared to that. detectable
after 5 million cycles.
IMPACT STRENGTH
The last item of this papev is impact strength of GRC rotorblades. This strength
being reasonable good the author wishes to point out that he does not intend to encourage
for any in flight inpact test. The service experience in this area is upto today:
Two mainrotors went through oak trees cutting off branches in excess'of 50 mm.
The rotorblades were thoroughly inspected and cleared for further service without repair
of the GRC.
A hovering BO-105drifted backward and the tailrotor impacted a MRB of an UH 1 D on
ground. All rotorblades engaged were severely damaged. The composite tailrotorblades
failed at the impact station only (Fig. 27). The remainders of the TRB provided suffi-
ci-nt thru3s to allc'w the pilot for a safe landing.
V ICONCLUSION
Weil designed and produced GRC is the best structural material for rotary wing
application known to the author. To.days experience indicates that there is no adverse
Jntluence on material properties by service environment.
The pure resisuance of GRC against rain- and sanderosion requires very carefully
designed erosion protection systems. Also the methods for repairing the prclective
coating need further improvement.
LITERATURE
1. Weiland, E.F.: "Development and Test of the BO-105 Rigid Rotor Helicopter",
J.American Helicopter Soc iol. 14, No. 1, January 1969
2. Barth, R.: "The Hingeless Rotor - A Concept to Increase Mission Effectiveness at
Reduced Costs". AGARD Annual Meetin4., September 1973
3. Reichert, G.:" The Impact of Helicopter Mission Spectra on Fatigue" in AGARD
Specialists Meeting on Helicopter Design Mission Load Spectra, AGARD-CP-206, 1976
4. Brunsch, K., Wackerle,P.M.: "Ballistic and Imnact Resistance of Composite Rotor-
blades",Second European Rotorcraft and Powered Lift Aircraft Forum,BUckeburg,
20. - 22.9.1976
5. Reichert, G., Weiland E.F.: "Long Term Experience with a Hingeless / Composite
Rotor" AGARD Flight Mechanics Panel Symposium on Rotorcraft Design, 16.-19.5.1977
6. Hoffrichter, J.S.:, "Evaluation of the Effect of Usage on Composite Main Rotor
Blades", Boeing Vertol Company, Philadelphia, 14.4.1978
i~
I,-
~ij
I.
i3* - .
18-7
IS ARTAICULATED ROTOR
CTEETERING ROTOR
BMRGENERAL ARRANGEMENT
VTA43 P~A
St
.t-i
I~ ~ ~-8-8 F ERSONTI SKN FARI
T 47EGASVLNAFIIH----~
"iI 450
FIMF2- COLY
RETH(PVC-HARD-OAM
+ + -t i
- 47~ 4-I4
720-
1900
EROSION STRIP
TiA 6V4
A hickSKIN,FABRIC
mm t450
POLYURETHANE-COATING
CORE, PVC-HARD-FOAM
SPAR, ROVING UNIDIRECTIONAL
FIBER E-i3LASS, AMINO-SILAN by WEIGHT
73.3
RESIN- EPOXY- BASED 2, yWIH
HARDENER: 4,4'-DIAMINO DIPHENYLMETHANJ
(AMINO-BASED)
EPOXY ADHESIVE
FILM FM 123-5.06
i4i
11104.128
.1filuments for
tc
spar only
"
FIG. 9: BO-105 BLADE SPAR BUILD UP
18-10
w
FIG, 12: IMPREGNATION QUALITY FIG.13: FIXTURE FOR RESIN CONTROL
LOWER PICTURE WITH OVEN
j R.
30.0
27.5-
00ce)-
.-- blad s after hs2er
-'.3ooo hrs)
17.5
n = 32
15.0 -
12.5 -
10,0 -
10ii0
N
FIG,15: BLADE SPAR SPECIMEN
SHEAR FATIGUE STRENGTH, A = 2/3, S = 50%, P 50%
N/MM 2 6 b,
450
300 -new
250-
J. 15o -*
10' 10'
.4p
18-12
BO-105 MAINROTORBLADE
FLAPP IN M014EN.
1.5 ,,
CHORDWI SE MOMENT '"
NNU.MBER
OF CYCLES
101 10,0
F T FATIGUE STRAIN
8_
2800 hrs S=50%
S , S=99,9%
3
!if|2800 hrs
MANEUVER (ESTIMATED)
0 077-
'NUMBER OF CCE
, 106 10 8
Il0=
i
111
tiLtJ -
-I Y
18-13
Mr
00000
00 2800 00 S=999%
3WO
-- MANEUVER LOADS
1106
00NUMBER OF CYLFS LEVEL FLIGHT I0
3700 hrs
S=50%
z
E 3700 hrs
! co-2800 hrs
4i-
MAXIMUM LEVELFLIGHT
105 10NUMBR OFCYCLES
106 10 10
K-1
- 18-14
J ~
n200
~~~43
% _____
_____ -
__
n 200
I n
>22
oo '3oo/hrs/
'a
C'.
18-16
it
"F!G. 26: PAINT ABRASION AFTER 1200 HRS GULF OF MEXICO (IMPROVED COATIN6)
I,
__ _ .
18-1'
35 3S
0
0
30 w030
C 2o
F 25 25FOE
i
c -
x -4i
is 15
10 I10
+ SHiFflWORK ESTII'4ATED
i !
itj
e.,
19-1
H. Benson Dexter
NASA Langley Research Center
Hampton, Virginia, U.S.A. 23665
SUMMARY
INTRODUCTION
Advanced composite aircraft structures have the potential to reduce airframe struc-
tural mass by 20 to 30 percent, reduce fuel consumption by 10 to 15 percent, and thus
reduce the direct operating ccsts to the airline operator. The U.S. Government and indus-
try have been developing advanced composites technology for the past 15 years. This tech-
nology has matured to the point where commercial aircraft manufacturers are starting to
make production commitments to this new class of materials.
The NASA Langley Reseacch Center has been instrumental in the development of advanced
composite structures primarily for commercial aircraft. Flight service evluation pro-
grams were initiated in the early 1970's to develop confidence in the long-termi durability
of advanced composites in actual aircraft operating environments. Concurrent with the
flight evaluation of structural composite components, NASA initiated programs to determine
the outdoor environmental effects, the effects of aircraft fuels and fluids, and the
effects of sustained stress on several thousand composite test coupons.
' In 1975, NASA initiated an extensive aircraft energy efficiency (ACEE) program to
improve the efficiency of commercial transporE aircraft through the development and appli-
cation of advanced technologies such as advanced composites. The objective of the ACEE
program is to accelerate the development of advanced technologies to the point where U.S.
commercial
production transport
aircraft. manufacturers can three
Six components, economically
secondaryincorporate
structures,the
andtechnology into their
three primary
bstructures,
M are presently under development.
The purpose of this paper is to outline some of the tASA/U.S. industry composite
programs and report on the service experience gained with numerous composite components
during the past decade.
Component Description
The composite components currently being evaluated in the NASA flight service
programs are shown in Figure 1. All components shown have been in service except for
the Bell 206L helicopter doors and fairing which are beirg fabricated. Details of the
current and planned flight service evaluation components are shown in Figure 2. The
first flight service program involved reinforcement of a U.S. Army CH-54B helicopter
tail cone with unidirectional boron/epoxy [1 and 2]. Boron/epoxy strips bonded to the
tail cone stringers increased the tail cone stiffness and reduced the structural mass
by 14 percent. Eighteen Kevlar/epoxy fairings are being evaluated on three Lockheed
L-1011 aircraft [7 and 8]. Solid laminate and Nomex honeycomb sandwich panels were
installed as a direct replacement for production fiberglass parts. An average mass
saving of 26.8 percent compared to the fiberglass panels was achieved.
The largest number of components is being evaluated on Boeing 737 aircraft [9 and 10).
One hundred and ecaht graphite/epoxy spoilers have been installed in sets of four by six
different commercial airlines on 27 aircraft for worldwide flight service. The spoilers
are constructed with graphite/epoxy upper and lower skins, corrosion resistant aluminum
honeycomb core, aluminum spar and hinge assemblies, and fiberglass closure ribs. The
finished spoilers are 35 percent composite and are 16.9 percent lighter than the standard
production aluminum spoilers. Two boron/epoxy-reinforced aluminum center-wing boxes are
in service with the U.S. Air Force on C-130 transport aircraft [3 to 6]. Boron/epoxy
bonded to the covers on both the wing planks and hat-section stringers resulted in an
increased fatigue life and a 10 percent mass saving. Three boron/aluminum aft pylon skin
panels are in service on three DC-10 aircraft [11]. These panels are subjected to
elevated temperatures and high-intensity acoustic loading as a result of their proximity
to the center engine. A mass saving of 27.3 percent compared to baseline titanium
was achieved.
Nine graphite/epoxy DC-10 upper aft rudders are in service on seven commercial
airlines [12 and 13]. These nulti-rib stiffened rudders weigh 32.9 percent less than
standard production aluminum rudders. An important aspect of the DC-10 graphite/epoxy
rudder program was the development of a cost competitive manufacturing procedure to
allow a single oven cure cycle for the total structural box. The process selected is
identified as the "trapped rubber" process and is shown in Figure 3. The process
begins with the layup and densification of the right and left side skin panels and pre-
forming the spars and ribs. The front spar web has lightening holes that permit the
spar to fit over internal metal mandrels of the tool. The mandrels are centered in the
cavity formed by adjacent ribs and the skins. Each mandrel is surrounded by carefully
sized blocks of silicone rubber that fill each cavity. Once all the ribs are located,
the outer skins are installed and heavy steel plates are attached to the tool to form
a closed system. The tool is rolled into an oven where temperature is increased at
specified rates and the rubber expands against the graphite/epoxy to develop the desired
cure pressures. Once the cure cycle is completed, the tools and silicone rubber are
removed and the finished structural box is ready for installation of standard aluminum
alloy hinges and actuator fittings and the standard glass/epoxy leading and trailing
edge members and tip assembly. Additional details on these NASA flight service components
can be found in [14].
The most recent flight service program involves installing 40 shipsets of Kevlar/
epoxy doors and fairings on Bell 206L commercial helicopters. The composite components
are being fabricated and flight service will begin in late 1980. Three design concepts -
stiffened foam sandwich, stiffened skin, and honeycomb sandwich - as shown in Figure 4,
will be evaluated. An overall mass saving of 25.9 percent compared to the production
components is estimated. The helicopters will operate in diverse environments in
Alaska, Canada, and the U.S. Gulf Coast. Coupons exposed to ground and flight
environments will be tested at specified intervals to determine effects of the various
helicopter operating environments on materizl strength. Also, selected components will
be removed from service aud tested to failure to compare residual strength with original
strength.
Sikorsky Aircraft has committed to extensive use of secondary and primary Kevlar/
epoxy and graphite/epoxy strvictural composites in their S-76 commercial helicopter.
NASA Langley Research Center and Sikursky are planning a joint program to determine the
long-term durability of composite components on the S-76. The objective of the NASA/
Sikorsky program is to determine the effects of realistic operational service environments
on typical composite helicopter structures. Fourteen tail rotors and four horizontal
stabilizers such as those shown in Figure 5 will be removed from helicopters after up
to 10 years of operational service. Tie tail rotor has a laminated graphite/epoxy spar
with a glass/epoxy skin. The horizontal stabilizer has a Kevlar/epoxy torque tube with
graphite/epoxy spar caps, full-depth honeycomb sandwich core, and Kevlar/epoxy skins.
The two composite components have a total mass of 24.7 kg. Static and fatigue tests will
be conducted on the components removed from service and the results will be compared
with baseline certification test results. In addition, several hundred composite
coupons exposed to the outdoor environment will be tested for comparison with the
component tes results.
19-3
- _.
___ ..
19-4
Component Description
S.
V Keviar/epoxy components account for approximately 45 percent of the wetted
external surface of the S-76. Applications include the horizontal stabilizer, main-
rotor blade-tip caps, canopy, radome, cabin doors, forward and aft engine fairings
above the cabin, landing gear and baggage compartment doors, and internal non-
structural items. Graphite/epoxy and Kevlar/epoxy hybrids are used in selected
components to improve strength and stiZfness. Several of the composite components
are shown in Figure 17. In most hybrid applications the graphite carries the primary
bending loads and the Kevlar carries the shear loads. An example is shown in Figure 18
for a horizontal stabilizer where uniaxial graphite/epoxy carries bending loads in spar
caps and cross-plied Kevlar/epoxy basically carries shear loads. The application of
graphite/epoxy in the tail rotor spar takes advantage of the superior fatigue
characteristics of this material for a primary, dynamic structural component. The
composite ccaponents have been flying for over 2 years on three S-76 prototypes with
no major malfunctions. The first commercia. delivery of the S-76 helicopter was i'
February 1979. As of December 1979, 27 S-76 helicopters have been deliveLed and a
total of 5450 successful flight hours have been accumulated on the composite compon.enIts.
Boeing-Vertol has an extensive development program underway to place Kevlar and
graphite composite components into production on Boeing 234 commercial helicopters.
Primary emphasis is on extending the range of the helicopter through rhe use of Kevlar-
graphite hybrid fuel pods as shown in Figure 19. Other graphite and Kevlar components
include doors, fairings, cabin floor, and support beams. The Boeing 234 helicopter is
scheIu]ed to enter commercial service in early 1981.
Bell Helicopter is using advanced composites in crew seats on the Bell 222
commercial helicopter. The seat shown in Figure 20 uses Kevlar skins and aluminum
honeycomb in sandwich construction for the bucket and a graphite/epoxy energy attenuator
tube as part of the seat support structure. Twelve model 222 helicopters have been
delivered with two lightweight energy-absorbing crew seats in each aircraft.
rudders. An additional 11 rudders have been manufactured under the ACEE program to
* develop more efficient manufacturing methods and to obtain quantitative cost data.
The rudders have been FAA certified and Douglas is considering fleet production of the
graphite/epoxy ruldders for new DC-10 aircraft.
The graphite/epoxy elevator for the B-727 is 0.9 m wide by S.C m long and has a
mass of 89.4 kg. The elevator design is dictated primarily by stiffness requirements
and makes efficient use of graphite/epoxy Nomex honeycomb sandwich skins to carry
normal pressure and in-plane shear loads. The upper and lower skins are attached to the
substructure with titanium fasteners. With this design, most of the interior ribs used
in the production elevators are eliminated and a mass saving of 23.6 percent has been
achieved. Since the elevator is mass-balanced, additional mass saving can be effected
through the use of graphite/epoxy. Eleven elevators have been fabricated to verify
cost projections and FAA certification has been received. Five shipsets of graphite/
epoxy elevators will be placed into airline service for evaluation. Four shipsets of
asse.bled graphite/epoxy elevators are shown in Figure 25. Boeing is considering
fabr, cation of up to 25 additional shipsets of elevators to establish manufacturing
learning curves.
Composite Primary Structures
Each of the three major U.S. commercial transport manufacturers are also under
NASA contract to design, fabricate, and test medium-sized primary composite components
as shown in Figure 26. Douglas and Lockheed will fabricate graphite/epoxy vertical fins
for the DC-10 and L-1011, respectively, and Boeing will build horizontal stabilizers
for the B-737. The ACEE composite primary components are nore complex than the
secondary components and thus present a greater design and manufacturing challenge.
Details of the three graphite/epoxy primary ccmponents are summarized in Figure 27.
The composite L-1011 vertical fin is about 2.7 m wide by 7.6 m long and has a mass
of 272.2 kg. The structural configuration consists of cocured hat-stiffened skins,
cocured I-beam stiffened front and rear spars, and multiple ribs. Using this design
approach a mass saving of 30.1 percent is projected. The tooling for the stiffeners
consists of inflatable silicone rubber bladders to provide internal pressure, exterior
caul plates, and a vacuum bag. The entire skin panel is cured in an autoclave under
pressure and elevated temperature. Upon completion of the cure, the inflatable bladders
are easily removed from the stiffener cavity as shown in Figure 28. Three L-1011 fins
will be fabricated to validate manufacturing cost projections.
The graphite/epoxy vertical stabilizer for the DC-10 is 2.4 m wide by 7.6 m long
and has a mass of 350.3 kg. The design configuration selected for the DC-10 vertical
fin consists of Nomex honeycomb sandwich skins, four I-beam spars with sine-wave webs,
and multiple sine-wave ribs. Titanium lug fittings are cocured into the spar caps to
provide root-end attachments. The ribs and spars are joined by adhesive bonded angle
clips and the skins are mechanically fastened to the substructure with titanium
bolts. A mass saving of 22.8 percent is projected using this design concept. Seven
DC-10 vertical stabilizers will be fabricated to obtain manufacturing cost data.
The graphite/epoxy horizontal stabilizer for the Boeing 737 aircraft is 1.2 m wide
*d by 5.2 m long and has a mass of 91.6 kg. The design selected consists of cocured
integrally stiffened skins, laminated front and rear spars with titanium lug attachments,
seven Nomex honeycomb ribs, two closure ribs, and a laminated trailing-edge
beam. inboard
The honeycomnb ribs have cutouts to allow passage of the cortinuous skin stiffeners
"as shown in Figure 29. Assembly of the structural box is accomplished "'th titanium
mechanical fasteners. With this design configuration, a mass saving of 22.9 percent is
projected. Eleven B-737 horizontal stabilizers will be fabricated to verify predicted
, I manufacturing learning curves.
Ancillary Test Plan
Numerous ancillary tests are being conducted by the ACEE contractors to verify
design procedures, provide design allowables data, and provide data in support of
".4 certification requirements. The tests vary in size and difficulty from small static
coupons to large combined-load subcomponents. Some of the tests are shown schematically
in Figute 30. As indicated, mechanical tests such as panel compression, root-end load
transfer, impact damage, rib-to-skin attachments, and lap joints are being conducted.
Most of these tests are conducted at room temperature or at elevated temperature with
*and without moisture cohditioning. Additional details of the NASA ACEE composite
structures program can be found in (15].
Advanced composites technology has been developed to the point wh-re U.S.
commercial aircraft manufacturers are starting to make production comm tnents to these
materials. Success of the flight service evaluation programs and technology developments
iunder the NASA ACEE composites program has led Boeing to plan extensive use of advanced
composites on their new B-767 aircraft. As shown in Figure 31, most of the control
K ;suzZices, "rcluding rudders, elevators, spoilers, and ailerons, will be built with
grephite/epoxy composites. Graphite-Revlar hybrids will be used in numerous structures
w!
19-7
such as leading and trailing edge panels, cowl components, landing gear doors, and
fairings. Replacement of standard fiberglass parts with Kevlar and graphite composites
will result in mass savings of up to 30 percent.
Lockheed has made extensive production conmnitments to Kevlar/epoxy secondary
structures on the L-1011-500 as discussed previously. Design studies are currently
underway to investigate potential graphite/epoxy structure applications on the L-1011
aircraft. Components being studied include the vertical fin, rudders, elevators,
ailerons, flaps, doors, and floor beams and posts. As shown in Figure 32, graphite/
epoxy applications being studied have a mass of 2560 kg which represents a 26 percent
mass saving compared to the current production components. At the conclusion of
these studies Lockheed will be in a position to make future production decisions
provided the composite components are economically feasible.
Douglas is also conducting studies to determine the economic feasibility of
committing to production of numerous composite components on their DC-10 aircraft. Some
of the potential components as shown in Figure 33 include graphite/epoxy vertical
stabilizers, rudders, elevators, spoilers, and ailerons. Other graphite/epoxy components
include doors, panels, beams, and nacelle structure. Potential Kevlar/epoxy componentE
include extensive fairings, tail cone, and nose cowl structure.
CONCLUDING REMARKS
Excellent experience has been achieved with approximately 200 composite components
in flight service for 2.5 million total component hours.
No significant degradation has been observed in residual strength of composite
components or environmental exposure specimens after 5 years service or exposure.
Technology is being developed that will lead to extensive production of advanced
composites in future aircraft.
Confidence in advanced composites technology is being developed to the extent
that commercial transport and helicopter manufacturers have made production commitments
to composites for selected components.
REFERENCES
}
Helicopter
1. Welge, R. T.: Application of Boron/Epoxy Reinforced Aluminum Stringer for the CH-54B
Tail Cone. Phase I: Design, Analysis, Fabrication, and Test.
Sikorsky Aircraft, United Aircraft Corporation, NASA CR-111929, July 1971.
2. Welge, R. T.: Application of Boron/Epoxy Reinforced Aluminum Stringers and Boron/
Epoxy Skid Gear for the CH-54B Helicopter Tail Cone. Phase II: Fabrication,
Inspection, and Flight Test. Sikorsky Aircraft, United Aircraft Corporation,
NASA CR-112101, July 1972.
3. Harvill, W. E.; Kays, A. 0.; Young, E. C.; and McGee, W. M.: Program for
Establishing Long-Time Flight Service Performance of Composite Materials in the
Center Wing Structure of C-130 Aircraft. Phase I - Advanced Development.
Lockheed-Georgia Company, NASA CR-.L]2120, September 1972.
K Flight
4.
Harvill, W. E.; Duhig, J. J.; and Spuncer, B. R.: Program for Establishing Long-Time
Service Performance of Composite Materials in the Center Wing Structure of
C-130 Aircraft.
S~CR-112272, April Phase
1973. II - Detailed Design. Lockheed-Georgia Company, NASA
5. Harvill, W. E.; and Kays, A. 0.: Program for Establishing Long-Time Flight
Service Performance of Composite Materials in the Center Wing Structure of C-130
Aircraft. Phase III- Fabrication. Lockheed-Georgia Company, NASA CR-13249b,
September 1974.
k' t
J. -
19-9
6. Harvill, W. E.; and Kizer, J. A.: Program for Establishing Long-Tire Flight
Service Performance of Composite Materials in the Center Wing Structure of
C-130 Aircraft. Phase IV - Ground/Flight Acceptance Tests. Lockheed-Georgia
Company, NASA CR-145043, September 1976.
7. Wooley, J. H.; Paschal, D. R.; and Crilly, E. R.: Flight Service Evaluation of
PRD-49/Epoxy Composite Panels in Wide-Bodied Commercial Transport Aircraft.
Lockheed-California Company, NASA CR-112250, March 1973.
8. Stone, R. H.: Flight Service Evaluation of Kevlar-49/Epoxy Composite Panels in
Wide-Bodied Commercial Transport Aircraft. Lockheed-California Company, NASA
CR-159071, March 1979.
Stoecklin, R~obert L.: A Study of the Effects of Long Term Ground and Flight
Environment Exposure on the Behavior of Graphite/Epoxy Spoilers - Manufacturing
and Test. The Boeing Company, NASA CR-132682, June 1975.
10. Stoecklin, Robert L.: 737 Graphite Composite Flight Spoiler Flight Service
Evaluation. Boeing Commercial Airplane Company, Fourth Annual Report, NASA
CR-158933, August 1978.
11. Elliot, S. Y.: Boron/Aluminum Skins for the DC-10 Aft Pylon. Douglas Aircraft
Corporation, NASA CR-132645, May 1975.
12. Lehman, George M., et al: Advanced Composite Rudders for DC-10 Aircraft - Design,
Manufacturing, and Ground Tests. McDonnell Douglas Corporation, NASA CR-145068,
November 1976.
13. Lehman, G. M.: Flight-Service Program for Advanced Composite Rudders on Transport
Aircraft. First Annual Summary Report, McDonnell Douglas Corporation, NASA
CR-145385, July 1977.
14. Heldenfels, R. R.: Recent NASA Progress in Composites. NASA TM X-72713,
August 1975.
15. Vosteen, Louis F.: Composite Structures for Commercial Transport Aircraft. NASA
Technical Memorandum 78730, June 1978.
16. Pride, Richard A.: Environmental Effects on Composites for Aircraft. NASA
Technical Memorandum 78716, May 1978.
ACKNOWLEDGEMENT
Portions of the data reported herein were obtained from U.S. commercial transport
aircraft and helicopter manufacturers. Several NASA researchers participated in
developing the data reported for the NASA composites program. The author gratefully
acknowledges these contributions.
it_
:N4
V. R
;Figure 1. NASA flight service composite components.
3;
7571 -7- IT,__E_ _ _ _ _
19-10
C~~PNET
CMOET
I COMPOSITE
MATERIAL
STRUCTURALBASELINECOMPOSITE
CONFIGURATION MASS, MASS.
MASS
SAVINGS.
______ I_____ kg kg
k9
RD
KW11OR
-4
FAIR NG
W-FRI
_Vt kEDPo
SHORIZONTAL STABILIZER
WUPAPIM4(EVLAR/EPOXY
SPAR
*KvLRtpoxysaw
~ NOMEX HONEYCOMB SAN~DWICH
L-1011
FAIRING PANELS 18 JANUARY 1973 17,718 282.870
737SPOILER 108 JULY 1973 17.
500 1,438.000
~* 1.5
RESIDUAL1.
STATIC
STRENGTHSAER AN FO
RATIO16 NEW SPOILERS
0 T30015209
0 T001544GRAPH ITEIEPOXY
0~ AS 13501 MATER IALS
0 1 2 3 4 5
CALENDER TIME IN SERVICE, yr
fFigure 7. Residual strength of graphite/epoxy spoilers.
19-12
5 YEARS SERVICE
GRAPHITEIEPOXY MOISTURE CONTENT,
PERCENT
T730015209 0.67
T300/2544 0.66
AS/3501 0.75
INSPECTIONINPTI"
COMPONENT INSPECTION
INTERVALC STATUS
months METHODS
-BA N m h VISUAL MINOR DISBONDS
CH-54BTAL _CONE _ 2 ULTRASONIC NO REPAIR REQUIRED
MINOR SURFACE
DC-10 AFT PYLON SKIN 12 VISUAL CORRO NNE
CORROSION ON ONE SKIN
DC-10 UPPER AFT RUDDER 3.12 VISUAL MINOR RIB-TO-SKIN
ULTRASONIC DISBOND ON TWO RUDDERS
',! ~~~~COMPOSITE
S
STRUCTURAL *BASELINE
MS, *COMPOSITE
MS, MASS
SVNS
COMPONENT MATERIAL CONFIGURATIONMASS. SAVINGS.
L-1011 COWL
DOOR GRAPHITE/EPOXY AL HIC SAND. 121.1 94.3 22.1
"L-1011 FLOOR POST GRAPH171EPOXY LAMINATED I-BEAM 1.7 1.3 23.5
,COMPONENTMASS PER AIRCRAFT
- i
19-!3
m
0.2
.6m
TYPICAL VERT.
STABILIZER
TRAILING EDGE
PANEL
~ I BEAM
CABIN -LOOR
-' NOSE
LANDING GEAR
DOOR
IIN
FARIG PANELS.EAD
WIN L. PNES
.TOTAL
KEV.R OOMPOSITE.
APPLICAIN 13 ~
Figre SS SAVED -366 kg
FiueI3 Lockheed L-1011 Kevlar/epoxy composite
.1 1$applications.
19-14
12
*KEVLAR-491EPOXV
- SKINS :,450 INOMEX
HONEYCOMB CORE)
SKINS 0 KEVLAR-491EPOXY
TORQUE TUBE 450
SKEVIAR LAMINATE
''.3 SNDIC
AIOSORP
~FULAOD
AND ~ ~ ~ HATC~
FULRODFARGG F
SUPPOR
REAMPOR BA
FUE
POD
S 19-16
RUBBER
D-IAPHRGM
SEAT
PAN SET
(UNDER
CUSHION) FIBERGLASS
KEVLAR
SANDIC
BUCKET
A'
GRAPHITE "FLEX-CORE'
TUBE
ATTENUATOR
~' OBJECTIVE
PROVI DE THE TECHNOLOGY AND CONFIDENCE SO THAT COMMERC IAL
TRANSPORT MANUFACTURERS CAN COMM ITTO PRODUCTION OF
COMPOSITES INTHEIR FUTURE AIRCRAFT:
SECONDARY STRUCTURE - 1980 TO 1985
PRIMARY STRUCTURE - 1985 - 1990
TECHNOLOGY CONFIDENCE
"* DESIGN CRITERIA, METHODS AND DATA 0 DURAB ILI1Y /WARRANTY
"* QUALIFIED DESIGN CONCEPTS 0 QUANTITY COST VERIFICATION
I *COST COMPETITIVC MANUFACTURING 0 FAA CERTIFICATION
IPROCESSES 0 AIRLINE ACCEPTANCE
Figure 21. NASA ACEE composite program.
A 2E4I
-Ra
c1- rev
t'i
o
Figure 26nAAAEcpitprmrsruue.
~ ..IRZ
COMPONENT ~~ET FI VET STB STAB.
L-01 D-1 -3
COPOEN
VET I ET TA.HrZ TB
S~SECTION VIEW
Ir
STIFFENER COCURED
WITH SKIN -
HILOK MECHANICAL
FASTENERS
Figure 29. Boeing 737 composite stabilizer assembly.
PRESUREIHEAR
SKIN PANEL TO RIB
STIFFENED SKIN PANEL ATTACHMENTISBO
COMPRESSION "
AND SHEAR SI ~
MECHANICAL JOINT
DISCONTINUOUS LAMINATE
SINBOARD ELEVATORS
LEADING
EDGE - UBOARD
IALILERONS
GP
Mi- NOORS ACCESS DOFORSIN
A -- 5w
19-20
MLG DOORS
NIG DOORS
IELEVATO
FLOOR BEAMS
RUDDER AND POSTS
ALRNFA
AILERONS
FLAPS--
TRAILING EDGEPANELS-':c V S
SPLEVAOS
BEAMSELVTR
FLOOR
LONG DUCT
NACELLES
SFigure 33.
NOSE GEAR DOOR
Potential composite structure applications
on Douglas DC-10.
q S-F
Is R0-STT-_L0
KEVLAR/F161
2- T30012544
LAMINATE KEVIARIF155
MASS
CHANGE.
PERCENT
AS13501
T30MJ052081
0 1 2 3 4 5
EXPOSURE TIME, yr
1.5
RESID UAL
FLEXURE
STRENGTH 0 T30/5209
RAIO0 T30I254
RATI 'Z ASI-3501
.5 0 T300W5208
V7 KEVLARIF161
0 1 2 3 4 5
EXPOSURE TIME. year
Figure 36. Residual flexure strength after worldwide exposure.
1.5
0 1 2 3 4 5
EXPOSURE TIME. yr
3ligure 37. Residual tensile~steghatrssandsrs udor-exposures;
T300/5208, [00-4090 aiaesrse t4 percent ultimate.
19-22
1.51
I
RES IDUAL
TENSILE
STRENGTH
RATIO
0 INITIAL AMBIENT AIR
.5 o JP-4
0 SKYDROL
A FUELIWATER
V FUELIAIR CYCLING
0 1 2 3 4 5
EXPOSURE TIME. yr
Figure 38. Effect of fuels and fluids on strength;
ST300/5209, 450 tensile specimens.
III
a .is O~ 8 oT3050 []
T8 o 8 MM
0a
8
- 80
0 8
I rso0 13I
6 .ma
J-- ~F
-- EXU~lE1MENHUUS 2l 2~
! " Figure 40. Composite mass change during flight exposure on Boeing 737-.
,1;:_ _ __
77
- 1' 7-77
20.1
Frank J. Fechek
Systems Support Division
Materials Laboratory
Air Force Wright Aeronautical Laboratories
Wright-Patterson AFB Ohio 45433
Summary: The chronology of the Air Force Materials Laboratory advanced composite development programs which
contributed to the current capability to use these materials in primary and secondary structures on high
performance military aircraft is described. Some of the concerns regarding the durability of these materials
in actual service environnents are being addressed by conducting a systematic, periodic non-destructive
evaluation of selected composite structures in operational service. Whereas visual and x-radiographic
inspection techniques have been found to be quite usable on composite structures in-the-field, itefficiencies
using available, portable ultrasonic inspection equipment in the field environment have-accentuated the need
for the development of a semi-automated, ultrasonic inspection system specifically designed to be compatible |o
with current, production composite aircraft structures. A system satisfying these needs has been shown to
be feasible and is now in the prototype stage of development.
The Air Force Materials Laboratory has been instrumental In the growth of the advanced composites
technology beginni g with the fiber and resin/prepreg development, continuing through the period of
structural design technique and configuration verification, and evolving into the manufa.,.uring processes
and cost effectiveness studies which eventually contributed to the commitment of this category of structure
into production use in advanced aircraft airfreses.
Beginning in 1961 the AFML sponsored the initial feasibility program to define the best chemical
processes to produce boron fibers suitable for reinforcement in composite structures. As these filaments
became available in sufficient quantities to verify the potential utility of high modulus reinforcements
in polymeric resins, effort was initiated in 1963 to also refine the processes to produce high strength,
high modulus graphite reinforcemeit f:bers in uselal quantities and at a reasonable cost. The availability
of sufficient quantities of these fibers from pilot plant development programs in the mid 1960's made
possible the initiation of structural hardware concept and development effcts beginning in 1964. At this
time, the United States Air Force, by instituting the Composite Structures Advanced Development Program at
the AFML, demonstrated its intention to solve the problems of transitioning this new technology from a
demonstrated-to-be-feasible, laboratory curiosity into a viable, integrated structural materials industry.
It was expected that the products of this new industry would make possible the redesigning of airframe
structures along with achieving equivalent or increased airframe structural component performance with
reduced weight. The approach taken to accomplish this transitioning goal was to select actual structural
components of current or anticipated aircraft systems as technology demonstration items and then, through
a series of design, fabrication, and test iterations, achieve a manufacturable component ready for flight
service evaluation. Some of these development programs resulted in ground testing of one-of-a-kind
structures, some resulted in linited production and in-service evaluation, and some have resulted in air-
frame structures which have been placed in routine production use. During the sixteen years since the
first structural demonstration item program was initiated in 1964, many material system combinations,
design concepts, structural configurations, and manufacturing techniques have been investigated. Table I
depicts the chronological progress of the development of the advanced composite technology and its
"introduction into United States military aircraft systems. This table shows that the mid to late 1960's
V were used to refine, test, and confirm the evolving design methods and manufacturing capability to use
composites ab skin members in full depth honeycomb structures. The overriding design-criteria during this
period was to duplicate the static and dynamic response of the equivalent metal design which thb-composite
part was to replace. This structural development period emphasized boron fibers, and although few articles
of any individual design were produced, most were demonstration-flight tested, and most exhibited ueight
savings in the 20% range.
The early 1970's were characterized by; A). high modulus graphite filaments becoming available to-the
designer, B). a series of programs directed at non-full depth honeycomb-structural configurations, C).
various attempts at non hand-layup manufacturing concepts, and, D). a continuing cautious attitude of the
designers abcur incorporating composites into the substructure portions of the composite assemblies. This
period continued to emphasize the manufacture of only a limited number of items of any single desrgn, with
no flight aemonstration of them except in-the case of the P-14, P-15, and F-ill production activity. In
a these latter instances, the technology of the 1960's, i.e. boron composite skins bonded to full-depth
aluminum honeycomb sandwich, was applied to production on a routine basis. The weight savings achieved by
these production items was as predicted based upon the experience gained in the development programs.
-jThe mid 1970's were characterized by a more aggressive and expanded use of this maturing technology-by
the airframe manufacturing community. Examples are; A). the many materials system-combinations attempted
L ~ 4 in the earlier periods had evolved to a few, widely used systems, B)., graphite fiber reinforced-applications
increased, C). composites as substructure were introduced, D). designs of increased sophistication were
attempted, and E). prototype aircraft used composites as primary bill-of-materials with no alternate metal
designs.
From the mid 1970's to the present, composites use has further increased in both numbers and types:of-
application. The designs selected during this period emphasize:- A). a more selective-use of reinforcement
materials for particular properties, B). a mixing of materials in a structure for both structuraloand
economic reasons, C). introducing and evolving toward mechanically fastened concepts, and D). aggressive
use of composites in strength critical designs.
3-A
A ~- --- _ _ _ _ _ _ _ --- -
20-2
Thus, the current status of the continuing evolution of these materials is; A). that their use as
secondary structu:es is routine, B). wing-type primary airframe loads and structures have been incorporated
into the most recent systems, and C). the technological basis of information and data have .:, obtained to
permit the future application of this t,chnology into fuselage type applications. Current develoraent
programs are addressing both the wing/fusel,.ie intersection requirements and the concerns about the long
term reliability and durability of these nuturials and design concepts.
Table !I contains a summary of those military aircraft systems which have incorporated the use of
advanced composites as airframe structure on a routine, production basis. This table shows that a consid-
erable amount of in-service experience has been achieved in the past ten years, measured either in calendar
time. or accumulated fleet or individual aircraft flighthours. Also shown in this table is the status of
one of the advanced development effort structures. It is shown here both for comparison purposes, because
it is graphite/epoxy, and because it is included in an in-service monitoring effort. Not included are-more
recent applications of this technology, where extensive amounts of-graphite/epoxy composites have been used.
The F-18 and the AV-8B are examples. These have been omitted because they are Lelatively new and have not
yet accumulated comparable fleet flighthours and environmental experience.
The AFML, in 1976, initiated a non-destructive inspection monitoring program of selected composite
structures in service with the objective of establishing some basic information on the durability of these
structures in an actual service environment. In this effort, standard NDE techniques are used to period.-
cally monitor the condition of a few selected structures. This effort is not intended to measure, nor
statistically indicate, the actual condition of the structures in the fleet at any point in time. Rather,
it is intended to establish a baseline of experience and data which could be used to plan future fleetwide
Inspection programs, should this become necessary, and/or to develop the information which would be used to
establish the criteria against which a portable, reliable, semd-automated in-the-field inspection system
could be developed. The composite structures on two aircraft systems were selected. These are the
composite structure portions of the empennage of the F-15 and the composite outer wing panels on several
A-7T) aircraft. The details of this program are indicated below.
Three locations were selected to visit annually to monitor F-15 aircraft. They are Edwards AFB,
California, Luke AFB, Arizona, and Langley AFB, Virginia. The two aircraft at Edwards APB were selected
because they are the second and eighth airframes manufactured in this program, thus they have been in the
service environment the longest calendar time (although they have not neceRsarily accumulated the most
flight hours). The three aircraft inspected annw.ally at Luke AFB are of interest primarily because they
are exposed to a desert environment, whereas the three aircraft inspected annually at Langley AFB experience
a typical seashore environment throughout the year. The structures inspected on the F-15's are the vertical
and horizontal stabilizer torque boxes and the rudders. All are full depth aluminum honeycomb construction
bonded to boron/epoxy skins. Typical construction is shown in Figure 1. Disbonds or delaminations in the
area of the skin to titanium step splice joints at the periphery if the skins or moisture incursion into
the honeycomb core aze the two primary effects expected upon the composite structures as a result of either
the ambient climatological environment or the flight loads. Thus, ultrasonic testing and x-radiography
were chosen as the in-the-field techniques to be used on this program. Because of the configuration of-the
skin to titanium close out joint in the torque boxes, a pulse-echo technique is very tedious. Because of
this, a through transmission technique has been adopted in which the signal generated passes through the
boron/titanium joint of the upper skin, through the honeycomb core, and again through a similar boron to
titanium joint before being received by another transducer. Failure of the generated signal to be received
is the indication of a disbond or delamination and such areas are further evaluated using pulse-echo
techniques with a single transducer. For the structural configurations tested to date, 0.375 inch, 2.25 MHz
transducers are used in conjuncLion with a Branson 301 portable ultrasonic instrument and have been found
to be adequate. A typical inspection involves completely inspecting the entire periphery of both~horizontal
stabilizers and the periphery of the lower two thirds of each vertical torque box using this technique.
The bonds are interrogated at one half inch centers and a total length of bonded steps of 1070 inches crn
typically be completed by a two person team in four hours.
In checking for moisture incursion, in order to restrict the inspection time required to a reasonable
amount, it has been assumed that the most likely areas which would inge t moistire would be those where the
skins were relatively thin. Thus, only the outboard 30 inches of each of the stabilizers and the complete
planform of each rudder are radiographed. Using a focal distance of 30 feet for the vertical stabilizers
and rudders and 13 feet for the horizontal stabiliz.-rs and exposing the st:Iictures to 125 kv at, 5 ma for
between 1.5 and 7.5 minutes (depending upon structure thickness and focal distance) results in obtaining
good quality radiographs using Kodak Industrex AA ready pack film. This procedure results in films which
contain a minimum amount of honeycomb image foldover so that either liquid nolsture or evidence of moisture
presence in the past (i.e. cell vall distortion) can be detected. Exposing 28 14x17 inch films per aircraft
takes a two person teaw 5 hours, including set up and take down time.
During the 43 months of this program, no delaminations or disbonds have been discovered in the
composite to titanium bonded joint areas. Likewise, only subtle evidence of moisture incursion into the
X- honeycomb has been discovered and this in an insignificant percentage of the area examined.
.: The physical configuration of the A-7D composite outer wing panel is shown in Figure 2. Twelve of
*
_ these structures were manufactured in 1976 as part of a manufacturing technology effort. The objective of
this in-service experience program is to monitor the performance of a representative graphite/epoxy
construction in both a desert and seashore environment with Air National Guard aircraft. Four-composite
outer wing panels (COWP's) have been stationed at Kirtland AFB New Mexico, and thren at McEntire ANGB,
Columbia, South Carolina.
I The A-7D in-service Inspection effort differs from the F:l5 program in the following ways: A). only
seven structures are included in the program (all the other wing tips in the fleet-are metal), B). the
P: 4-1 COWP's are a hybrid construction including B/E, Ti, fiberglas, and Gr/E. C). the-assembly is sandwich skin
mechanically fastened and bonded to a composite rib/spar substructure, and D). x-radiography is unsuitable
Sas an inspection technique and the return signals from this-complex atructural arrangement makes the use of
203-3
ultrasonic techniques of marginal utility. Thus, visual inspection is the primary technique used on these
structures, a boroscope being used on the interior of the structure. A Branson 301 portable ultrasonic
unit in the pulse-echo mode, using a 0.375 inch, 2.25 MHz transducer has-been used to-check skin to sub-
strucr-re suspect areas. To date, since mid-1977, the semi-annual inspections have recorded no major
degradation attributabie to the in-service environment. The following have been experienced, however; A).
some minor cracks in cut edges of the panels at the radii of access parel hole frames, and B). some minor
damage where edge damage impact can occur. No structural disbonds and no damage attributable to in-service
abuse have been discovered. Additionally, no extra maintenance has been required on these structures (in
fact, they require less than the comparable metal structures) and no repair actions have been required.
In summary, the performance of these graphite/epoxy structures in service has been quite satisfactory.
An edditional experiment is being conducted in conjunction with the COWP's on the A-7D's. This
involves installing unstressed 4x6 inch panels on racks on the underside of COWP access doors. The objec-
tive of this experiment is to monitor the change in physical and mechanical properties which these materials
experience when exposed to real time environmetital exposure identical to that of the wing structures. The
parameter, being compared are geographical location (i.e. low vs high ambient humidity), coated vs uncoated,
and various material 3ystems. Table III sumnmrizes some of the properties measured during the first 24
months of this ccmtinuing effort.
A general summary of the observations resulting from these in-service inspection monitoring efforts
is: A). the boron/epoxy structures appear to be quite inert to their service environment. They have not
experienced any degradation resulting from either their flight loads or their typical ground handling
environments and procedures. The structural bonds in the skin to titanium step splice appear to be
similarily degradation resistant. Only an extremely few occurrences of possible moisture entry and minimal
core damage haje been observed. Of these, the seashore stationed items acquired most of the occurrences,
although a small percentage were discovered on aircraft stationed in the desert. Recorded moisture in
seashore aircraft "vanished" after these aircraft were transferred to the desert location. B). the
graphite/epoxy structures generally likewise seem to be inert to the loads and ground handling environment;
However, in the case of the specific structures inspected, some minor delaminations have been experienced
and some crack:.ng at cut edge corners has been discovered. It is not evident whether these few isolated
events are problems peculiar to this individual design or a basic materials system inadequacy, C). in-
service ultrasonic inspection of aircraft using cff the shelf equipment and techniques is time-consuming
and prohibitivcly expensive if anything other than small surface areas are to be inspected, D). observed
changes in "flen" size using ultrasonic testing in very subjective and operator dependent, especially when
attempting to compare findings from one year to the next, E). off the shelf NDE ultrasonic equipment
suitable to. in the field use is not capable of providing a printnd permanent record usable for future
Sreference.
Thus, a secondary conclusion from this effort has been that a need exists for a more efficient NDE
tool to inspect this category of aircraft structure in-the-field.
Toward satisfying this need, the AFML, in 1978, initiated a progran to define the feasibility of, and-
then develop, a portable, reliable in-service ultrasonic -nspection system capable of in-the-field use.
This system, as conceived, would; A). enable the operator tj rapidly scan the structure, B). automatically
detect flaws by zomparing the ultrasonic signal against some predefintd standard, C). store in computer
memory the flaw locations, D). indicate, in real time, where the flaw Iccations were and which portlons of
the structure had not yet been interrogated, E). provide, in near real time, a permanent record printout of
scanned area, flaw locations, and a computed flaw depth and, F). upon operator command, digitize the RF
signal of the flaws and store in computer memory. This program is called "In Service Inspection System
(ISIS)." The bantic functional elements of the ISIS concept are shown in Figure 3. These elements have
been organized ui-to various assemblies as follows; A). Transducer Asser'bly--this hand held unit is composed
of the transducer, the solid or liquid delay line, a couplant dispensor, a sound emitter, push button to
activate the wave form digitizer, and a signal loss alarm light, B). Position Sensing Assembly--this
apparatus positions the two, point source microphones which send the signals to the graphpen digitizer
which in turn computes in rectanguiar coordinates the position of the transducer within the 24x24 inch
operational inspection area and transmits this information to the CPU, D). Ultrasonic Assembly--currently
a Sonic Mark IV (Sonic Instruments, Inc, Trenton, N.J.) has been modified to incorporate a dual flaw gate,
a flaw amplitude digitizer, and waveform digitizer and is used for these functions, E). Data Acquisition/
Processor Assembly-a Motorola MC 6800 Microprocessor with a 16 K byte ROM and a 48 byte RAM is used as
j the CPU for this system. Tle ROM contains the system level software. It cuntains 16 parallel I/0 channels.
F). Data Display Assembly--a Techtronics TEK 4025 Graphics Display is used to input instructions to the
system and display in near real time the ultrasonic scan completion status and flaw locations on the CRT.
It can also be used to display digitized waveforms, and G). Data Recording Assembly--this Hewett Packard
9876 A Printer/Plotter will provide permanent copies of C scan recordings, digitized RF waveforms, and
coordinate/amplitude listings.
The ISIS block diagram, Figure 4 shows the relationships of the various functions. The operational
3'
procedure of the unit is; A). the outline of the area to be interrogated is projected on the CRT display
based upon information fed to the CPU by placing the transducer assembly at the corners of the perimeter
of the area of interest and activating a button-on this assembly, B). each possible data point 0.1 inch-by
0.1 inch in the area to be inspected is illuminated by a dot.on the CRT, C). the reflected ultrasonic signjal-
is identified with its location, then analyzed against predefined criteria and recorded in memory, D). if
the interrogated area is a flaw, the display dot increases in brightness and if it is-not, the-dot ii-
extinguished, thus, whtn the completa surface has been scanned, only-flaw locations are illuminated, E).
the flaw areas are !.utomatically recorded and the printer plotter will produce a permanent copy of the
results of the scan showing those areas which have and have not been scanned, F). for a given flaw, the
operator can activate the waveform digitizer and record and print the digitized representation of-the RF
waveform, G). this information can also be used to automatically compute the depth of the flaw by ratioing
the flaw reflection time to the difference between the near side reflection time anu the far .side reflection-
time.
M.
Sa
W-i
Ia.
I 5
tA
tt _
20-5
Table I
Chronology of Advanced Composite Aircraft Applications
46.. 04
IWC
Fiber O
BEx 5505 __ _ _ _ _ _ _
Horizontal
Stabilizer
Wing Pivot
71
72
F-5
7-111 II I
EdgeFlap300
Epxy & lStfn'd
HTS
3002
Boron
F. D.
Sandwich
Panel
Panel_
X
N
Yes
_ _ _
824
_ _
2
_ _ _ _
25
_
j
-. Fitting Doubler 5505 -- - ... ..
S =Underwing 72 F-Ill AS-i F. D. j ~ I Yes 824
miril
r ng 3 5 01 Sa nd wich - -__ _ _ __ _ _ _ _ _
I i- Landing Gear 72 F-5 HTS F. D. N 9
Wheel Door 3002 Sandwich - I
- F. D. -
1Win Full depth honeycob sandwich.
* 20-6
Table I (Cont'd)
0 4
e0 0 ,
0> 1 0
0-
00 A
A1 Fairing
Landing Gear
Door
75 F-14
5505/3501
AS
3501
Sandwich
F. D.
Sandwich
xIIe
I ______I__
____________
7 8
Wing Trailing 75 B-1 AS F. D. No
Edge Flap 3501-5 Sandwich -IX
Weapons 75 B-1 AS F. D. X jx No 8
SBay Door 3501-5 Sandwich C' ' C
!VAvionics 75 B-1 AS F. D. No- 2
Bay Door 3501-5 Sandwich C '
Leading 75 B-1 Boron/AS N
Edge Slat 5505/3501-5 Sandwich x xN
Forward 76 F-16 Kevlar/T300 Stfn'd Panel xNo 1 21
Fuselage 5208 ____________
-i X
20-7
Table II
Summary
Military Aircraft using Composite Structures in Production
Date - First Ylight Aug '731 Dec '70 July '72 Feb '74 June '771
Total Aircraft 4122 349/803 483/213 100 74
Cumulative Airframe 377,200 255,000 224,600 9854 4324
Flighthours
Maximum Flighthours 20851 1469 1200 892 7271
per Airframe
Cumulative Composite 1,508,800 510,000 1,347,6005 49,270 4324
Component Flighthours
Composite Component(s) wing root pivot horizontcl horizontal horizontal and Composite
fitting doubler stabilizer and vertical vertical outer
underwing fairing stabilizers stabilizers, wing
and rudder vert. leading pauel
edge and rudder
1 with composite structures
2 currently flying with composite doublers
t 3 Foreign sales - no data
i 4 only aircraft with COWP's
5 excluding speedbrake
TaLle III
j
I Polurethane
Paint
one surface
6-8
11-14
18-20
23-26
0.06
0.10
0.16
0.25
0.38
0.52
0.63
0.70
0.46
0.81
0.78
1.01
Polurethane 6-8 0.02 0.35 0.44
Paint 11-14 0.06 0.46 0.78
both surfaces 18-20 0.09 0.56 0.70
23-26 0.16 0.68 0.94
Aluminum Foil 6-8 0.11 2 0.30
N (bonded) 11-14 0.11 0.59
one surface 18-20 0.17 0.67
V 23-26 0.13 0.91
Aluminum Foil 6-8 0.02 2 0.25
(bonded) 11-14 0.02 0.51
7 both surfaces 18-20 0.03 0.66
23-26 0.02 -0.88
F
S2
1 rooftop exposure (in shade)
specimens lost in aircraft crash
InI
4V
7 20-8
I- SECT IO
Fiur-
Tyia P-5Cntuto
SETO A- A
GRAHIEI.4
HYBRI
SKINPANEL
-EOX
GRPIE-PX SUSRUTR
HOEYOM SKN.SASADRB
INUTUTR
AU 'AESET
COPOIT
AECTIONU fl-A
IIA
IIIA
FigreI
zt:/
4-DCmoieotrWn ae
ISIS -CONCEPT
HIGH RESOLUTION SONIC MICROPHONE POSITION MICROPROCESSORIGRAPHICS
ULTRASONIC TECHNIOUE SENSING SYSTEM DISPLAY INTEGRATION
R
SENSO
IGN~~t MICROP'HONES GAHC
NARROWINTRFAEE
SNAPLSE
IDTH
NIGHROWINERFAETASUE
SI N LPU
LSE WIR
TH
K YBOAR
t
MICROPRIOCESSOR
Vt Figure 3
Functional Elements of ISIS
k
RA
INEFC 7
R"C
4IAI
6
IRA"-
I-~~~ pIROROEtO
~RINTit
PtIGOTE
4MAGNETI
INTEFAC
ISIS BloOSIiTION
INERAC ULTASNI
r 020-10
CALIBRATION LIST
REFERENCE STD, NO. H/A
.TRANSDUCER SER. NO. V103
FREQUENCY (hHZ) $
LOGIC HEG
GAIN SETTING
COARSE 80
FINEC 8
DEC SLOPE0!%:
DEC RANGE
FRONT SURF GATE (USEC/10) 24
MlAX STEP (PLIES) 2
OPERATING MODE I
--
Figure 5
/: A
ii
U. S. NAVY SERVICE EXPERIENCE WITH ADVANCED COMPOSITES
by
A. Somoroff, H. Dubberly, J. H. McGinn, and M. Tarricone
Naval Air Systems Command, Washington, D. C. 20361
and
A. Manno, Naval Air Development Center, Warminster, Pa. 18974
1. INTRODUCTION
The U. S. Navy is entering into a period of expanding composites use in major air-
craft systems. These application committments, made several years ago, were based upon
clear advantages in structural efficiency or cost and expectations of good service dura-
bility. The well knowcn corrosion resistance and generally outstanding material fatigue
behavior of composites coupled with limited but largely positive actual service experi-
ences supported these expectations. Of course, it has been recognized that composites
exhibit environmental sensitivities and other characteristics which differ markedly from
conventional metallic behavior and which must be accommodated in design and service sup-
port. As might be expected, present designs reflect generally conservative judgments on
expected lifetime environmental exposures. As actual long term operational experience
accumulates, the distillate of this experience can be expected to shape the nature and
extent of future composite applications as well as the supporting design development and
fleet maintenance practices.
i The F-14 horizontal stabilizer was the first production application of composites in
flight critical structure. As shown in Figure (1), the stabilizer is composed of
boron/epoxy skins over full depth aluminum honeycomb core. The skins vary from 56
plys at the root, where they are bonded to a titanium splice plate, to 7 plys at the
) tip.
A review of all maintenance data from Mid-1976 to late 1979 shows the~majority-of
the damage to be cosmetic in nature such as repair of the paint, corrosion of the
aluminum lightning protection strips and dents in the leading/trailing edges -and -tip
.assembly. Minor damage is repaired by squadron personnel while major punctures-are
returned to the depot. Typical examples of minor and major damage are shown in Fig-
ures (2) and (3), respectively. This review also shows that an average of four sta-
bilizers are damaged beyond the fleet repair capabiliti4s (2" diameter hole)- each
21-2
year. Approximately 30% of thep- units are repairable by the depot. Since the
damage is being caused by handling as opposed to a design or manufacturing problem
these trends are expected to continue.
A program which will increase the depot repair capability to deal with 6 in. dia-
meter holes is about half way completed. Initial tcits have shown that the repair
concept being employed will recover in excess of design oltimate load capability.
Further fatigue and environmental testing will begin soon. The test specimens and
test set-up are shown in Figures (4) and (5), respectively.
The Navy recently removed from service a horizontal stabilizer that had seen approx-
imately five years of service use in the fleet. This stabilizer is being subjected
to an extensive laboratory examination to determine the effects of service use
(real-time loading, temperature, humidity, etc.). The examination consists of both
non-destructive and destructive testing including testing approximately 200 speci-
mens cutout from the skins. While the study is not yet complete the results to date
have shown minor damage such as dents and areas of honeycomb core corrosion. Tests
of moisture content and mechanical properties are becoming available. In thinner
laminate sections (20 plies or less), beneath the lightning protection strips, resin
a moisture content varies from approximately 1.5 to 2,1 percent of resin weight. For
thicker sections (40 to 48 plies) resin moisture content was about I perceht. Be-
F tween the lightning protection strips, resin moisture content was found to be some-
what less. No significant differences were found between laminate specimens taken
from the upper and lower surfaces of the stabilizer. Measurements of the resin con-
tent of all laminate specimens ranged from 25 to 29 percent by weight and were with-
M in original manufacturing requirements.
The existing H-46 metal rotor blade has been a source of continuing maintenance ex-
pense. The major cause of H-46 metal blade failures in service were undetected
manufacturing defects, atmospheric and fretting corrosion, mechanical damage, and
gunfire. The lack of tolerance to defects resulted in the necessity for many blade
inspections in service and a high rate of return to depots for inspection and re-
pair. A sealed spar pressure crack detection system was developed and retrofitted
1to to mnito
monitor conditions in serice.Malunctons
i conitins
service. f this
Malfunctions of ths cack
crack etecionsystemade
detection s madded
'3 to an already low reliability level and increased the unscheduled removal rates.
r The attrition rate of metal blades due to corrosion is substantial.
r IThe
S3.2 Composite Blade Description
H-46 composite rotor blade is shown in Figure (6). The rotor blade consists of
a fiberglass "D" spar terminating in a two-pin, four-lug, wrap-around root end re-
f tention. Graphite is located along the spar vertical heel web-to increase flapwise
stiffness and section torsional stiffness. A titanium/nickle leading edge nose cap
provides all-weather erosion capability for the blades, some torsional stiffness,
lightning protection, and FOD protection for the composite leading edge. Titanium
was selected in this application because of its formability and fatigue strength-
- which makes it compatible with the fiberglass/graphite.
components. During the course of the development, considerable effort was expended
to establish production techniques that were cost effective. The composite blade
program consisted of three phases as follows:
This phase consisted of preliminary blade design, production blade design, tool
design and fabrication; fabrication of tool proover and 14 test blades; reli, bility
development/qualification via fatigue and whirl tests; and a flight test progr.,m.
This phase consisted of tooling and minor design updating from the initial blade
fabrication, fabrication of a pilot-production quantity of 72 blades, a relia-
bility demonstration test utilizing ten flight aircraft in a closed cycle con-
tractor supported program, and a flight demonstration of blade interchangeability.
The Phase II [Figure (8)) reliability testing was a key "Lement of the development
plan and was conducted from May 1978 to Hay 1979 under a variety of fleet opera-
ting conditions by four squadrons at four different locations. Sixty blades were
utilized and a total of 18,658.8 blade flight hours were accumulated with no re-
turns to depot required. The average hours per blade during the evaluation de-
monstration was 311 hours and the high time blades-six accumulated 778 hours. All
blades continue in operation. Figure (9) provides data showing the cumulative
blade hours for the program. Fourteen accountable repairs were accomplished by
organizational or intermediate level maintenance. The problem areas (i.e. de-ice
blanket wiring harness, inboard and outboard rubber closure rib) were identified
early and timely corrective actions (i.e. manufacturing changes, design improve-
ments, maintenance procedures) were initiated. It should be noted that the struc-
tural integrity of the blade was unaffected by any of the problems encountered
and that no significant design defects were manifested. Approximately 200 pilots
flew the H-46 with fiberglass blades during the evaluation. The vast majority of
pilots rated flight performance good to excellent for all maneuvers.
The development and implementation of practical repair concepts iab a primary ob-
jective of the maintainability program. Field level repairs and riiterials were
identified aud developed. Sixteen typical repairs, Figure (10), were made to an
outboard blade fatigue test specimen prior to test. Fatigue testing confirmed
the adequacy of the repairs.
When fiberglass rotor blades were reinstalled following removal for maintenance,
no new tracking adjustments were required. No difficulties with rotor blade
interchangeability have been reported.
The experience from the service reliability program was used to maximize the
likelihood that problems were identified and corrected prior to full scale pro-
duction commitments. Approval for service use was granted in November 1979 and
production rotor blades are being delivered for use on fleet aircraft.
Hr- Production and delivery of blades for the fleet is now underway with 339 -de-
livered through the end of 1979. Nearly 50,000 rotor bladu flight hours have been-
t Z accumulated through the end of 1979. Service experience continues to be excellent.
Rate production will continue for several years until metal blade attrition re-
quirements are satisfied.
Although service experience is very limited, experience to date indicates the Ha-46
composite blade may well provide a solution to every major problem existing-with
4no the H-46 metal rotor blade. Cost studies indicate that composite blades will cost
more t-an metal blades in production. With reduced spares -requiremets, seem-
ingly excellent repairability in the field and higher reliatility in service, the
admittedly still limited information supports the likelihood of a much lower over-
I rall cost of ownership.
8A. Two YAV-8B aircraft have been built as flight test vehicles in support of the
AV-8B Advanced Harrier full scale development aircraft.
The 28 foot span wing torque box is a multi-spar design. The upper and lower covers
are simple monolithic solid laminates of AS/3501-6 unidirectional graphite/epoxy,
continuous from tip-zo-tip and variable in thickness and orientation to provide for
the many different loading conditions. The substru-ture is primarily sinewave con-
struction fabricated with T300/3501-6 woven graphite,0epoxy broadgoods. Metal fit-
tings have been used at all significant load introduction points and where substan-
tial out-of-plane loads exist. Assembly of the torque box is accomplished by bolts
and nuts and no secondary bonding has been attempted.
The flaps and ailerons are of similar construction as the torque box except that
channel type spares and ribs are used in most locations due to the lesser thickness
of these components. Due to its very shallow thickness, the flap door is a full
depth honeycomb design, with aluminum honeycomb core and graphite/epoxy skins. The
outrigger and overwing fairings are relatively lightly loaded components and are
fabricated as syntactic sandwich construction and are further stiffened with either
cocured hat section stiffeners or separately cured stiffeners, subsequently bolted
on. Three wings have been built, two for flight aircraft and one for static struc-
tural testing which has been successfully completed.
| Well into the assembly of the wing, it was discovered that additional shear
VF strength was required in a portion of one rib. Since the lower skin was already
F attached to this member, removal was impractical. An auxiliary rib section was fab-
S .... ricated and .nstalled
been installed on a metal rib. to the existing rib much as a doubler would have
adjacent
- During the flight test program, there were several instances of fuel leaks-at
fasteners through the graphite/epoxy skins, Figure (13). None of these leaks were
Z
Me|
-4 -t
A porosity condition on a fuel tank spar web which did not infringe on structural
adequacy but which allowed fuel seepage was readily resolved. A partial vacuum, was
applied to the fuel tank and epoxy was painted on the external surface of the spar
webs, which was then drawn into the porous areas - resolving the problem.
- The Lift Improvement Devices (LIDS) fence and the graphite/epoxy strakes, Figure
(14), are subject to impact damage from objects blown about during vertical land-
ings and takeoffs and during power nozzle braking. No such damage was observed.
- Flight testing and service experience of the YAV-BB to date has revealed no prob-
lems attributable to the composite materials.
The S-3A spoilers are surfaces hinged off the rear spar of the wing. They function
as roll control devices and speedbrakes. Each wing half has four spoilers - three
upper and one lower. The lower spoiler, which was selected for composite applica-
tion, is located on the wing underside as shown in Figure (15), and is a simple
beam supported by two hinge fittings and positioned by a push rod attached to each
- fitting. The lower sp,.iler planform is quadrilateral about 85 in. long, 8 in. wide
at the outboard end, and 15 in. wide at the inboard end.
A low-temperature cure graphite reinforced epoxy system was used in order to reduce
processing costs and to utilize a high strength/high peal adhesive for bonding. The
material system selected was 5209/T300. Adhesives used were METALBOND M-1113 and 3M
AF126 film. These materials have a 260*F during temperature with service appliea-
tion temperature limits of 180F which is suitable for the spoiler environment.
= ILightning protection is provided by an aluiinum wire mesh-screen bonded to the ex-
ternal skin during cure of the laminate.
Construction details being evaluated in service are low temperature cure laminates ae
- 1j and adhesive, thin skin honeycomb construction, potted insert attachments and wire
mesh lightning protection.
ri:i Fourteen ship sets of composite spoilers were fabricatee. Ten were placed in ser-
S jvice and four were held as spares. One ship set was installed on a test aircraft in
April, 1975, to undergo flight testing at the Naval Air Test Center, Patuxent River-,
MD and has remained on the aircraft since then. The remaining ine ship sets were
installed in February, 1977. Six were assigned to a Pacific Coast squadion and
three were assigned to an Atlantic Coast squadron. Accompanying the spoilers were
oNDI standards and inspection/repair instructions. As of 30 November 1979, a total
Ii of 331 service
assigned to the months andCoast
Pacific 10,060 flight have
squadron hoursbeen
havedeployed
been accumulated. Spoilers-
aboard carriers.
_4
21-6
One ship set of spoilers was removed after 30 months of service for laboratory eval-
uation. The spoilers were first inspected using ultrasonic C-scan and X-ray. -No
delaminations, unbonded areas, or signs of water infiltration were detected. One
spoiler was static tested identically to previous tests performed during preflight
ground testing. Deflection characteristics, failure load and failure location-4ere
essentially the same as obtained in previous tests indicating no detrimental struc4
tural effects occurred as a result of the 30 months servide environment. The, spoiler
will be dissected, inspected, cut into small specimens, and tested to. determine any
changes in material properties. Other spoilers will remain in service for a period
of five years during which time several ship sets will be removed and laboratory
tests performed.
This program, described in detail in reference (3), has been devoted to testijig of
composites in a carrier environment through installation of graphite-epoxy access
doors on F-4J aircraft. The principal objectives of this program were to verify
corrosion prevention methods used in graphite-epoxy/metal combinations, and deter-
mine the amount and rate of water absorption in graphite/epoxy composites or naval
aircraft. Materials selected and corrosion prevention methods considered are thoses
which are intended for application on the F-18 and AV-8B.
Two types of F-4J doors were selected for replacement with graphite/epoxy test doors:
1) an electrical access door, approximately 19 x 15 in.; and 2) a-fuel access do-or,
8 in. in diameter. The locations of these doors on the F-4J are shown in Figure
(19). As door orientation when installed on the aircraft could affect water reten-
tion, a vertical and a horiztntal door were selected.
Eight graphite-epoxy test doors, four electrical access and four fuel accessi were
fabricated. These doors contained various test specimens which were attached--td-the
inside of the doors. The test specimens consisted of various combinations of graph-
ite/epoxy, aluminum (core and sheet), titanium, and Nomex core either adhesively
bonded and/or mechanically fastened to their i.nterior surfaces. Each door also- con-
tained twenty-four graphite-epoxy tabs for m..isture content monitoring. Twelve were
bare and twelve were painted. All tabs were twelve plys thick.
During February and March of 1977, the eight test doors were installed on-four
carrier-based F-4J aircraft, one set (an electrical and a fuel access door) pnr- air-
craft. Two of these aircraft were assigned to the Atlantic Fleet and- two, to-the.
Pacific Fleet.
During service, these doors were inspected bimonthly. Door attachments were ixamined
for evidence of corrosion and one painted and one unpainted tab were removed- and-
sent to MCAIR, St. Louis for moisture content measurements. No corrosion was found
on any attachments or fasteners during two years of monitoring. The moisture-co-
tent was consistently higher for painted than for unpainted tabs and cycled witht-the
seasons, being higher in the summer when it was hot and humid and lower in thei-vin-
L I 6. SUMMARY
REFERENCES
1
I- Weinberger, Somoroff and-Riley, "U.S. Certification of-Composite-'Wings for the-
F-18 and Advanced Harrier Aircraft', -AGARD Report N 660, ..April 1977-.-
S BORONIEXPOXY
ALUMINUM LEADING
EDGE /
/1 TITANIUM
MAIN SPLICE Al INTERCOSTAL
~ STEEL
-,ROOT RIB
/ AND HORN
~/OUTER INNER
// BEARINO\--
~ "~ BEARING
ALUMINUM HtC
/
1STUBS
BORONIEPOXY
SHEAR
VINTER ~ AFCOT-
-ix
- ~ -~ - -~-~17
g a
i REPAIR SITE-TERRACES
t~
21-9
I
w
_ _____fil
_P=3
--P-
if~_
- nm
Wa~w h
9L
Cy 1977 1978 1979 1980 1981 1982 1983 -1984 1985 1986
ASUP7 ASU V
PHASE 3-PRODUCTION
FY 7826 12f77 6 262
F" 79420 3115179 309 M83
FY o.526 240 286
CUMULATIVE DELIVERIES (14) 77 339 648 1051 1454 1857 2260 2663 2839
1978 1979
J___ MIAIMIJI JIAISIOINI D J FFLIGHT
iF _ HOtURIS
ME J~~I
FIM JI )JAI SI OIJ I I F
BLADE DEL. RATE 1
I
11210 10 11216 1121121121 1 1 i -M
A/C BLADE
CUM. NO.OF BLADES 12 12 12 24 30 42 54 66 1 CUM. HRS. CUM. HRS.
TRACKMASTER 2 6
- A/C NO. 1 o',MX.1 c 564.2 3,385.2
A/C NO. 2 ' 6
-~i 1. 4,666.2
A/C NO. 3 C-I16 6 533.2- 3,199.2
A/C NO. 4 C-1i6 6 537.8 3,226.3
Itji ~A/C NO. 5 ~i ~ 66 140.0 -408.0-
A/C NO. 6 5 154.2 925.V2
E A/C NO. 7 -MAG;26 6 119.9 71790.4
SET .
SHIP V 6. "
SPARE BLADES
TOTALS j 72 3,169.818,658.
. INDICATES POINTS IN THi. PROGRAM WHERE THE SCHEDULED - -
;.:. NOT BLADE INSPECTIONI' Q'F CONDUCTED.
.50,000
45,000
40,000
35,0100
U, 30,000
25,000
co 20,00
15,000
RELIABILITY
10000TEST PROGRAM
10,00 -0PHASE 0-
COMPLETE 18,610 HOURS
5,000
j19791
0
1 1978
VA (. AW
LEADING____() A A
EDGE4 I
,I~ TRALINO
EDGE
~ii A _ _ _
10 Ol0IOWA
AhUN 1(051.09 09
lom
Il X9 11009*10 U lUCIA
All 90104M0 ftit0f I.% toCIA
vol00 'a s 0 ATO W1 slowiACI
FIGURE
0-11.46COMPOSIEIBLAPEALPAIR
21-12
Aluminum EDGE
FIE
FIUR BOTEREPIROLING
q2 IN SI
EDGE __
Legend:~_
FIGIGURE11-BABOLBECOREPAITEPARTS
I Iw
l~a ---
21-13
Fluorosilicone Ti Sealing
Al ompundHead-Bot
Leading Edge O-Ring
Formi-in-Place Seant,
t-Skin Monolithic
Panel GrinEp
AuxSpr-
I ~~Sine Wave Front SparGIp
I Monolithic Skin Panel
rim
IIi
Cavit
Fec
4
II
PL 45)45 PLY
WW 2 PLY DOUBLER
POTTED INSERT
GRAPHITE EPOXY
Tal
FIGURE17-ENITRAC DAMAGE ON S41A COMPOSITE SPOILER
21-16
MOI
* PAINTED
~1oUNPAJilTD ACS
1977_________ 1973
FUE
ACESSDOOR
c~i51.0
a A J J AIS0N0JFNANS
IMAM S1A 60NID - FlJu
EXPOSE DU RAPPORTEUR
par
J-M.Fehrenbach
Centre D'Essais Aeronautique de Toulouse (C.E.A.T.)
23 Avenue Henri Guillaumnet
31I056 Toulouse Cedex
France
Cinq papiers ant Stg prfisentfis dana ]e cadre de cette session relative 1 l'expfi-
rience en service. Les deux premiers concernent lea pales d'hMlieoptares tandis que-les trois
autres prdaentent un panorama asses gdndral de l'expfirience en service aux-Etats-Unis sur avionsA
civils et militaires, avec une allusion plus limitfie au cas des hilicoptares.
Le prdsent rapport a pour objet de souligner les tb~mes qui ant ftig relevds comme
particuligrement significatifs au cours de cette session, en tenant compte au mieux des fillments
apportfis par Is discussion qui a suivi.
Westland consid~re que l'axpirience acquise devrait permettre A l'avenir une r6-
duction des facteurs de sdcurit:6 actuellement utiliads pour tenir coapte-de Is dispersion des
caractdriatiques du matdriau.
Une question a 6ti posge concernant Is corrfilation entre lea rdsultats d'essais
sur 6ldments et sur tronqoas de pales. 11 arrive en effet que lea modes de rupture, dana-lea
conditions d'essai rfialisges en labo soient diffdrents de ceux de piaces rfielles. Toutefois, lea
-lea difficultfis de rdparation par l'utilisaceur pour lea pi~ces reistivement sophistiqudes que
sont lea pales d'hdlicoptares.
Ce travail, commencdi das los annfiea 1970, reprdsente sujourd'hui une importante
expeirience accumulfie (200 6ldments en vol, totalisant 2,5 mlin er opsnY
V.1
ME mlin erexcmoatY
US
* R5-2
L'analyse des rdsultats a mis en.6vidence le faible niveau de digradation consta-
tfiau terme de plusietirs annndes en service. Le riniveau de confiance" obtenti progressivement
amane A Stendre A une part plus grande de la structure l'usage des-matdriaux composites. Cette
ddmarche se concrfitise dana le cadre du remplacement du mdtallique par du composite our des
avions anciens ou par une u~tilisation extensive de la technologie composite pour l'ensemble
des fillments mobiles ou dfimontables pour des projets nouveaux (B. 767).
Des do-nAges ont 6tg constatfis en service, des r~parations ant fitf mises au point
et on a constatd gingralement Ia faible Eivolution dana le temps des digradations observdes.
M- Le niveau de confiance obtenu a conduit a fitudier l'emploi des composites pour
des structures de fuselages (F 5, F 16).
j L'US NAVY a 6goalment ddveloppi 1'usage des composites our sea appareilo, avec
des El6ments importanta our F 14 et F 18 at le remplacement des pales mnitalliques du H 46.
D I SC US SIO N
a -La contr8le de Ia qualitfi a 6t6 soulignS A plusietirs reprises comme 6tant un problame im-
F portant pour l'emploi des composites dano lea structures prinaires. Ce problame rav~t piti-
sietiro aspects
I -Ia qualit6 doit 8tra assurlie I tous leas miveaux de la production at bien sflr contralge
avec soin compte tenti de la sensibiliti des caractdiristiques A la pr~cision du procasous
P-l industrial
leniveati d'exigence quant aux performances des moyans de contr6le dripend largement du
-& caract~re critique otinon critique des piaces concernfes. Pnur des piaces dimensionnfies
AWW: en-rigidit6 (Eillments mobiles our avions), le seuil de d~tection de dfifauts paut Stre
t relativement large. 11 n'en sot pas de mgme pour des piacas vitalas dimerisionnfies am
rfisistance.
A.
xM.
-RS-3
c -Llexpdrience
3
ME~I
4
REPORT DOCUMENTATION PAGE
".Recipient's Reference 2.Originator's Reference 3.Further Reference 4.Security Classification
of Document
AGARD-CP-288 ISBN 92-835-0273-6 UNCLASSIFIED
S.Originator Advisory Group for Aerospace Research and Development
North Atlantic Treaty Organization
7 rue Ancelle, 92200 Neuilly sur Seine, France
6.Title
EFFECT OF SERVICE ENVIRONMENT ON COMPOSITE-MATERIALS
7.Presented at
the 50th Meeting of the AGARD Structures and Materials Panel, held in Athens,
Greece on 14- 1 April 1980.
8.Author(s)/Editor(s) 9.Date
Various August 1980
10.Author's/Editor's Address 11.Pages
Various 338
14.Abstract
. This meeting consisted of 21 presentations divided into 5 sessions:
- Physico-Chemical Effects of Environment
- Environment Superimposed on Stressing
- Mechanical Effects and Hazards
- Physical Hazards (lightning, rain erosion...)
Case Studies of Service Experience.
j qN The effects of humidity and of impacts, particularly when tests were made in compression,
were emphasized. Damage due to lightning appeared to be less serious than had previously
been supposed.
Some theories were propounded for the calculation of erosion, for the prediction of properties'
J4 iin a vacuum and for fatigue damage.
Service experience, on the other hand, proves to be extremely favourable and has not shown,
- =:up until now, any evidence of disturbing phenom~ena.
4p-M
AI
vt
co ba
c w
0 0 2 0
a8 ~-0 r
Z Z - Z0
a 0-- E-
r)4 En 0A O~z o(
o0 > 0 - 08 >o 0 P 0
ON( *~ DWa -c
0 bo
0 0
U ~ u 0 - V, -Su
'm0.0
>1 1 00 .!
>C 000
lU 2U 0
00t 00 0 ~ OUU
00 - O',.
~ E C-3C.
0
0~' 0 0
00a -- :3
II~i 0
-9 2. * i
0~
i0
~tDel
0
0. 0
(nCA0 40.
0,
1.) 0 -I
. 4r..
A 4-1
4.0 ~ .t-0- .
m -c
C 04
4)4 0 4 44
(4- be 44 0
0 4** 0 zcy
.4.0..00
040
oo. cm0 0.
0,0605
4t. 0 5-' .
0 * o" 640 4
0
U ro 0 .o
4 )
g 4) - 0 en)
00 r. . ( eq~-
0 )-0
07 x ~ O)4 2' s4
0- 04 o
m..OE c
.0 ~A.0 0c Q. co 0.0 )
. <A
r 4)4.>
as 04 0 CJ E co 00
0.0
- C
4
~ .o
r. 00 E
r
V
.2 ca 14 c -
LI)4
0 o-
J j 0 1) .0
4))
0. 4 4) 145
0 0. 0
.o.
cq 04)0 0<r 0 94) <
EA0 0 ta
NATO OTAN
It U OF
DISTRIBUTION
AGARDB UNCLASSIFIED
TION S
7 RUE ANCELLE 9
. 2200 NEUILLY-SUR-SEINE PUBLICATIONS
FRANCE FRANCEAGARD
Telephone 745.08.10 - Telex 610176
AGARD does NOT hold stocks of AGARD publications at the above address for general distribution. Initial distribution of AGARD
publications is made to AGARD Member Nations through the following National Distribution Centres. Further copies are sometimes
available from these Centres, but if not may be purchased in Microfiche or Photocopy form from the Purchase Agencies listed below.
NATIONAL DISTRIBUTION CENTRES
BELGIUM ITALY
Coordonnateur AGARD - VSL Aeronautica Militate
Etat-Major de la Force Adrienne Ufficio del Delegato Nazionale alI'AGARD
Quartier Reine Elisabeth 3, Piazzale Adenauer
Rue d'Evere, 1140 Bruxelles Roma/EUR
CANADA LUXEMBOURG
Defence Science Information Services See Belgium
Department of National Defence NETHERLANDS
SOttawa, Ontario KI A OK2 Netherlands Delegation to AGARD
DENMARK National Aerospace Laboratory, NLR
Danish Defence Research Board P.O. Box 126
sterbrogades Kaseme 2600 A.C. Delft
Copenhagen 0 NORWAY
FRANCE Norwegian Defence Research Establishment
O.N.E.R.A. (Direction) Main Libraty
29 Avenue de la Division Leclerc P.O. Box 25
92320 Chatillon sous Bagneux N.2007 Kjeller
GERMANY PORTUGAL
Fachinformationszentrum Energie, Direc~io do Serviqo de Material
Physik, Mathematik GmbH da Forca Aerea
Kemforschungszentrum Rua da Escola Politicnica 42
D-7514 Eggenstein-Leopoldshafen 2 Lisboa
TREAttn: AGARD National Delegate
GREECE
Hellenic Air Force General Staff TURKEY
Research and Development Directorate Department 3f Research and Development (ARGE)
Holargos, Athens Ministry of National Defence, Ankaa
UNITED KINGDOM
ICEL AND Defence Research Information Centre
Director of Aviation Station Square House
c/o Flugrad St. Mary Cray
Reykjavik Orpington, Kent BR5 3RE
UNITED STATES
National Aeronautics and Space Administration (NASA)
Langley Field, Virginia 23365
Attn: Report Distribution and Storage Unit
THE UNITED STATES NATIONAL DISTRIBUTION CENTRE (NASA) DOES NOT HOLD
* STOCKS OF AGARD PUBLICATIONS, AND APPLICATIONS FOR COPIES SHOULD BE MADE
DIRECT TO THE NATIONAL TECHNICAL INFORMATION SERVICE (NTIS) AT THE ADDRESS BELOW.
PURCHASE AGENCIES
Microfiche or Photocopy Microfiche Microfiche
National Technfical Space Documentation Service Technology Reports
Information Service (NTIS) European Space Agency Centre (DTI)
5285 Port Royal Road 10, rue Mario Nikis Station Square House
Springfieid 75015 Paris, France St. Mary Cray
VirginiaS2216!, USA .England Orpir.gton, Kent BR5 3RF
Requests for microfiche or photocopies of AGARD documents should include the AGARD serial number, title, author or editor, antd
publication date. Requests to NTIS shoula include the NASA accession report number. Full bibliographical references and abstracts
of AGARD publications are given in the following journals:
Si Scientific and Technical Aerospace Reports (STAR) Government Reports Announcements (GRA)
published by NASA Scientific and Technical published by the National Technical
Information Facility Infoitnation Services, Springfield
Post Office Box 8757 Virginia 22161, USA
Baltimcre/Washington International Airport
Maryland 21240; USA